University of Glasgow: Precision Medicine Projects

University of Glasgow: Precision Medicine Projects

Below you will find an exciting and diverse range of available 18/19 MRC DTP in Precision Medicine projects. The projects listed below are all based at the University of Glasgow. 

Please note that all applications have to go via the University of Edinburgh - even projects that are based at the University of Glasgow. Information on 'How to Apply' is available here.  

Applying joint longitudinal and survival models to identify trial subgroups with distinct outcome change to inform prognosis

Dr Jim Lewsey, Institute of Health & Wellbeing, University of Glasgow
Dr Pardeep Jhund, Institute of Cardiovascular and Medical Sciences, University of Glasgow
Dr David Kao, University of Colorado School of Medicine

PhD Project Summary

Generally, in randomised trials the focus of analysis is estimating treatment differences for a primary outcome measure at a pre-specified follow-up time point. The repeated clinical, symptom-related and quality of life measurements made at study visits at different time points are often not fully modelled which is both an inefficient use of the data collected, and could lead to important prognostic information being overlooked.

Over the last decade, there has been growing interest in the joint modelling of longitudinal and survival data, with both methodological advances and increased application in clinical research. The advantages of joint modelling are that it reduces bias (due to informative censoring of the longitudinal measurements when a patient dies or is lost to follow-up), increases statistical power and, perhaps most importantly, allows for different types of statistical inference: 1) what is the association between the longitudinal process and survival (e.g. is improving quality of life over time associated with increased survival?); 2) treatment effects on the longitudinal process; 3) treatment effects on survival [1].

A good candidate model for the longitudinal process is the latent class growth model because the measurement times do not have to be equally spaced, both measured and unmeasured covariates are accounted for, and measurement errors are appropriately dealt with [2]. The supervisory team have experience and expertise in latent class modelling of a single time point with heart failure patients [3], as well as repeated measures over time in clinical trials using traditional survival and regression analysis [4]. This research is a natural extension of prior work. Identifying distinct recovery/improvement trajectories among heart failure trial patients, and what baseline patient characteristics are associated with this, would be valuable for clinical decision-making. Further, this research will explore whether treatment influences the trajectories, and, in turn, the association with survival. Such findings identified for population subgroups would have particular importance for trials that produced overall null effects.

The data for this research will be drawn from the BioLINCC data repository ( In scoping work, we have identified many chronic (e.g. BEST, DIG, HF-ACTION, TOPCAT, SOLVD, SCD-HeFT) and acute (HFN-CARRESS, HFN-DOSE AHF, HFN-RELAX, HFN-ESCAPE) heart failure trial data sets which have appropriate longitudinal and survival outcome data as well as information on other clinical and demographic information measured at baseline.

• carry out a systematic review of applications of joint modelling of longitudinal and survival data in heart failure populations
• identify suitable trial data sets from BioLINCC
• where appropriate, combine trial data sets into analysis-ready data sets
• carry out joint modelling using latent class growth models
• develop user-friendly interfaces to illustrate the utility and implications of the research to clinical audiences

Training outcomes:
• key data science skills of database management and data manipulation using software favoured by experts (SQL, R, Python and Stata)
• student will be trained in variable-centred (e.g. regression-type models) and person-centred (e.g. latent class growth analysis) statistical methods
• application development to illustrate model results in a user-friendly way (e.g. using Shiny package in R)

1) J Ibrahim, H Chu, L Chen. Basic concepts and methods for joint models of longitudinal and survival data. Journal of Clinical Oncology 2010; 28: 2796-2801.
2) P Wang, W Shen, M Boye. Joint modeling of longitudinal outcomes and survival using latent growth modeling approach in a mesothelioma trial. Health Serv Outcomes Res Method 2012; 12: 182-199.
3) D Kao, J Lewsey, I Anand, et al. Characterization of Subgroups of Heart Failure Patients with Preserved Ejection Fraction with Possible Implications for Prognosis. European Journal of Heart Failure 2015; 17: 925-935.
4) Jhund, P. S., Anand, I. S., Komajda, M et al. Changes in N-terminal pro-B-type natriuretic peptide levels and outcomes in heart failure with preserved ejection fraction: an analysis of the I-Preserve study. European Journal of Heart Failure, 2015; 17(8): 809–17.


Defining the relationship between rheumatoid arthritis, comorbidity, and adverse health-related outcomes: A precision medicine approach

Dr Stefan Siebert, Institute of Infection, Immunity and Inflammation, University of Glasgow
Dr Bhautesh Jani, Institute of Health & Wellbeing, University of Glasgow
Prof Frances Mair, Institute of Health & Wellbeing, University of Glasgow
Dr Barbara Nicholl, Institute of Health & Wellbeing, University of Glasgow

PhD Project Summary

Rheumatoid arthritis (RA) is a chronic inflammatory arthritis characterized by joint pain, swelling, and damage. RA is associated with significant comorbidities and increased mortality. Cardiovascular disease is a well-recognized comorbidity; an individual with RA is almost twice as likely to develop a heart problem as someone without RA. It is also recognised that RA and cardiovascular disease share key inflammatory processes, with much research focused on this area(1). The impact of other common physical and mental health morbidities in RA is less well known. Objective markers of inflammation may be of less relative importance than subjective factors related to wellbeing(2). This impact will vary by individual and will be the result of interaction between biological, lifestyle, and environmental factors. Risk stratification of patients with RA is not currently part of routine clinical practice; understanding these relationships will open the door to precision medicine methods to explore the potential stratification of people with RA, in order to better identify and treat patients at high risk to help prevent adverse health-related outcomes.

Cohort studies provide an excellent opportunity to explore these relationships. This project will involve working with Scottish, UK, and Swedish datasets, which include sociodemographic, biologic, morbidity, and lifestyle data on general population (UK Biobank) and RA specific cohorts (SERA and Swedish Arthritis cohorts) with linkage to health-related outcomes data. UK Biobank includes over 500,000 adults aged 40-70 years of which over 5,500 (1.3%) report RA(3). The SERA cohort is a Scotland-wide inception cohort of >1000 people with RA and undifferentiated inflammatory arthritis, incorporating patient and clinician-reported data2. The Clinical Epidemiology and Rheumatology Sections of the Karolinska Institute (KI), Stockholm, holds or has access to a number of large RA inception, disease, and treatment cohorts(4).

This project will give the successful student the opportunity to develop models of risk stratification in large cohorts of patients with established RA (UK Biobank and KI), which can then be tested in separate inception cohorts for RA (SERA and KI) to see how predicted outcomes perform compared to observed outcomes.

This project provides the opportunity to develop models of risk stratification in large cohorts of patients with established RA (UK Biobank and KI), which can then be tested in separate inception cohorts for RA (SERA and KI) to see how predicted outcomes perform compared to observed outcomes. Research questions:
1. What is known about predictors of mortality and adverse health-related outcomes among people with RA and comorbidity?
2. How do demographic, lifestyle factors, morbidity measures, and routine clinical blood tests mediate the relationship between individuals with RA and adverse health-related outcomes (death, cancer, hospitalisation and cardiac events)?
3. Is it possible to develop a predictive algorithm to identify individuals with RA and comorbidity at high risk of adverse outcome?

The proposed aims will include a specific focus on multimorbidity, to evaluate whether risk of adverse outcomes for people with RA is associated with specific comorbidities, clusters of comorbidities, or the cumulative number of comorbidities, and if inclusion of these provides additional utility beyond other risk factors identified by previous literature.

Training outcomes:
This PhD project, under the guidance of an experienced and multidisciplinary supervisory team, will provide the student with a solid grounding in systematic review, population health, and data science approaches, including in silico-type modelling approaches. The student will receive specific training and develop competencies in epidemiology and statistical analysis using complex and innovative methods. Generic research and communication skills will also be developed. The student would visit and be hosted by KI for a few months to undertake a component of analysis (see letter of support).

1. Holmqvist M, Ljung L, Askling J. Acute coronary syndrome in new-onset rheumatoid arthritis: a population-based nationwide cohort study of time trends in risks and excess risks. Ann Rheum Dis. 2017 Jul 14. pii: annrheumdis-2016-211066. doi: 10.1136/annrheumdis-2016-211066.
2. Dale J, Paterson C, Tierney A, Ralston SH, Reid DM, Basu N et al. The Scottish Early Rheumatoid Arthritis (SERA) Study: an inception cohort and biobank. BMC Musculoskeletal Disorders. 2016 Nov 9;17(1):461. doi: 10.1186/s12891-016-1318-y
3. Siebert S, Lyall DM, Mackay DF, et al. Characteristics of rheumatoid arthritis and its association with major comorbid conditions: cross-sectional study of 502 649 UK Biobank participants. RMD Open 2016;2:e000267. doi: 10.1136/rmdopen-2016-000267
4. Askling J, Fored CM, Geborek P, et al. Swedish registers to examine drug safety and clinical issues in RA. Ann Rheum Dis. 2006;65:707-712. Doi: 10.1136/ard.2005.045872

Demonstrating value for high priced novel treatments in oncology in an era of precision medicine: methodological and policy tools

Prof Andrew Briggs, Institute of Health and Wellbeing, University of Glasgow
Dr Kathleen Boyd, Institute of Health and Wellbeing, University of Glasgow
Dr Peter Bach, enter for Health Policy & Outcomes, Memorial Sloan Kettering Cancer Center

PhD Project Summary

The cost of cancer drugs is increasing at an alarming fashion. Work from Memorial Sloan Kettering Cancer Center in New York has shown that the median monthly cost of a course of cancer treatment at product launch has grown exponentially over the past 40 years . In the UK, the public watch-dog, the National Institute for Health and Care Excellence (NICE) determines value for products entering the UK market and provides recommendations for their use within the NHS. However, due to perceived lack of access to innovative oncology treatments, previous governments have implemented a ‘Cancer Drugs Fund’ (CDF) to provide access to novel cancer treatments that NICE has not approved for use. In the first five years of this initiative, the fund was overspent triggering an National Audit Office investigation . In attempt to recify the issues, the management of the failing CDF was recently returned to NICE. This all points to the specific challenges that expensive cancer therapy places on health systems worldwide and explains the recent interest in ‘value frameworks’ for cancer therapies even in the US where the formal assessment of value, as a policy tool, has been rare historically.

Much attention has focused recently on the use of precision or stratified medicine to provide “the right medication, to the right patient, at the right price” suggesting the possibility of increasing value of better targeted therapy within a precision medicine framework. Yet the reality is somewhat different, with smaller targeted groups of treatments leading to increased costs such that it is the manufacturers that capture the economic benefits of improved outcomes for patients rather than the health systems.

All this points to an urgent need for principled, rigorous evaluation of the comparative effectiveness and value of novel oncology products to guide prioritisation and decision-making withn health systems. Improving policy making in this area requires a good grounding in sound methodological tools of comparative effectiveness in the face of many methodological challenges. These challenges include the need to extrapolate from surrogate endpoints, such as markers of response or progression to overall survival, while simultaneously controlling for important clinical and biomarkers, and the possibility of patient crossover to active treatment. Further methodological development of Outcomes-Based Contracting and Stratified Medicine (including the posiblle interactions between these approaches) offers the potential to maximize health for a cancer patient population within a given budget constraint while ensuring value for the payer. The aim of this PhD is to explore these methodological issues surrounding the assessment of value in cancer care while also showing the translational aspects that are required in order to drive value in the context of cancer care delivery. The recent collaboration between Health Economics and Health Technology Assessment at University of Glasgow (led by Professor Andrew Briggs) and the Center for Health Policy and Outcomes (led by Dr Peter Bach) offers a unique opportunity for the student to learn from the perspectives of two international figures in two very different health system contexts.

1. To develop methodological tools around comparative assessment of value for cancer products involving the need to simultaneously address the need for extrapolation and allowance for biomarkers in the context of evidence that is incomplete, involves crossover and indirect comparisons.
2. To show how these methods can be applied to develop policy relevant tools that drive value, including outcomes-based contracting and stratified medicine
3. To understand the challenges faced by different health systems internationally, principally the UK and the US, when assessing and applying concepts of value-based treatment for their citizens.
4. To capture, in health terms, the benefits of applying a rigorous health economics approach to the definition and application of value in oncology.

Training outcomes:
The student will benefit from the in-house training activities available at HEHTA including:

• Decision Modelling Methods for Health Economic Evaluation
• Systematic Review and Evidence Synthesis (to include network meta analysis and indirect treatment estimates of comparative effectiveness)

In addition, the student will benefit from an extended placement at MSKCC and be able to access appropriate institutional data frm both MSKCC and University of Glasgow to support their PhD studies.

References:, accessed August 10th, 2017, accessed August 10th, 2017

Determining the role of Hedgehog signalling and anti-apoptotic proteins in conferring chemoresistance in AML

Prof Mhairi Copland, Paul O’Gorman Leukaemia Research Centre, Institute of Cancer Sciences, University of Glasgow
Dr Helen Wheadon, Paul O’Gorman Leukaemia Research Centre, Institute of Cancer Sciences, University of Glasgow
Prof Christopher Gregory, MRC Centre Inflammatory Research, University of Edinburgh

PhD Project Summary

Acute myeloid leukaemia (AML) is a somatic stem cell disorder, and the most common malignant myeloid disorder in adults, with an annual incidence of approximately 3.8 per 100,000. Untreated AML typically results in bone marrow failure, leading to fatal infection, bleeding, or organ infiltration, within 1 year of diagnosis, but often within weeks to months. Treatment and prognosis can vary significantly between patients, overall survival varying from 20-47%, depending on subtype, mutational status and age. Treatment is associated with considerable morbidity and mortality however, and a cure for the majority of adults remains elusive. Additional/alternative therapies are therefore a focus of intense research.

Hedgehog (Hh) is a ligand-dependent signalling pathway acting in concert with other stimuli in the specialised microenvironment of the leukaemia stem cell (LSC) niche in the bone marrow (1). In AML, aberrant Hh signalling has been linked to drug resistance with inhibition of the pathway restoring chemosensitivity; the bone marrow microenvironment is critical in this. Very recently, early phase clinical trials have demonstrated the efficacy of inhibiting Hh signalling using the SMO antagonist Glasdegib (PF-04449913) alone and in combination with chemotherapy in AML (2).

Studies have highlighted that BCL-2 and MCL-1 are over-expressed in AML (3), with expression linked to chemoresistance and poorer survival. The anti-apoptotic molecule BCL-2 is directly regulated by the Hh pathway components, GLI-1 and GLI-2. Clinical trials are currently exploring the efficacy of the BCL-2 inhibitor venetoclax, alone or in combination with chemotherapy. MCL-1 and PARP inhibitors (4) also show promise in preclinical models of AML, but have not been fully explored in LSC and bone marrow microenvironment models.

1. To characterise the chemoprotective effect of the bone marrow microenvironment and its effect on apoptosis in AML;
2. To define how Hh pathway inhibition alters apoptosis in AML;
3. To evaluate the transcriptomic, proteomic and functional effects of modulating apoptosis using (i) the BCL-2 inhibitor, Venetoclax, (ii) the MCL-1 inhibitor S63845, and (iii) the PARP inhibitor olapraib alone and in combination with either conventional chemotherapy (cytarabine or azacytidine) or SMO inhibition (glasdegib) in different molecular subtypes of AML;
4. To explore the importance of drug sequencing in maximizing the efficacy of these novel combinations again AML CSCs.

Training outcomes:
The project will provide exposure to a broad array of cellular and molecular techniques including: processing and culturing cell lines and primary cells from patients; in vitro co-culture systems; flow cytometry; Western blotting; RNA extraction and RT-PCR techniques; Q-PCR and mini-array Taqman-based PCR, including single cell PCR; RNAseq sample preparation and data analysis; cellular drug treatment; cell cycle, proliferation and apoptosis assays; and NSG xenograft assays. In addition, the student will be fully trained in dealing with the safety and ethical issues of working with human samples. Thus the student will become highly skilled in a range of in vitro techniques and in vivo models in the pre-clinical development of novel cancer therapies.

Excellent support is available from all supervisors who are in the laboratory on a daily basis and also their post-docs. The student will learn about liaising with the NHS and tissue bio-banking for research purposes and related ethical issues. In addition, the research team has  strong links with industry for example, from working with pharmaceutical partners in the preclinical development of novel small molecule inhibitors. Thus the student will gain experience in collaborating with industry and the issues surrounding this, e.g. intellectual property, funding, clinical trial development.

Based at the Paul O'Gorman Leukaemia Research Centre, the student will attend regular national/international seminars, laboratory meetings, work-in-progress meetings and journal clubs, enabling the student to gain the expertise and confidence to present their data at future meetings and conferences. They will also be given the opportunity to attend and present their results at National and International meetings. Beginning with informal lab meetings and culminating in international conferences, the student will be trained in rigorous thinking about their research and given the experience and confidence of presenting their work at the biggest forums.

1. Campbell V, and Copland M, 2015. Hedgehog signaling in cancer stem cells: a focus on hematological cancers. Stem Cells Cloning 8:27-38.
2. Cortes J, et al, 2016. A Phase 2 Randomized Study of Low Dose Ara-C with or without Glasdegib (PF-04449913) in Untreated Patients with Acute Myeloid Leukemia or High-Risk Myelodysplastic Syndrome. Blood 128: abstract 99.
3. Teh TC, et al, 2017. Enhancing venetoclax activity in acute myeloid leukemia by co-targeting MCL1. Leukemia. 2017 Jul 28. doi: 10.1038/leu.2017.243. [Epub ahead of print].
4. Esposito MT, et al, 2015. Synthetic lethal targeting of oncogenic transcription factors in acute leukemia by PARP inhibitors. Nature Med 21:1481-90.

Developing Multi-Modal Biomarkers for Prediction of Psychosis and Mental Health Outcomes in At-Risk Populations

Dr Peter J Uhlhaas, Institute of Neuroscience and Psychology, University of Glasgow
Prof Stephen Lawrie, Deptartment of Psychiatry, University of Edinburgh

PhD Project Summary

Schizophrenia (ScZ) is typically a debilitating mental illness with a lifetime prevalence of approximately 1% which leads to enormous economical and social costs. One critical factor in potentially improving the outcome would be the identification of individuals at high-risk for the development of ScZ to allow the possibility to intervene prior to the full manifestation of the syndrome. Evidence suggests that ScZ is preceded by a prodromal phase involving attenuated, psychotic symptoms of up to 5 years which are associated with a reduction in brain tissue and cognitive deficits (Fusar-Poli et al., 2013).

While screening procedures are characterized by sufficient diagnostic accuracy to detect at-risk individuals, these approaches are currently not sensitive and specific enough to predict psychosis-risk on an individual level, a key objective for early intervention research. Accordingly, biomarkers may be required to boost prediction and would likely allow insights into the underlying neurobiology of the at-risk state. The search for pathophysiological mechanisms and biomarkers is, however, complicated by the fact that CHR-participants have highly variable outcomes. This is highlighted by recent findings that CHR-participants who do not make the transition to psychosis develop a range of psychiatric disorders, including affective and personality disorders.

The current project aims to develop a biomarker for stratification and risk prediction of mental health outcomes and psychosis in a large sample of participants (n = 150) meeting clinical-high risk criteria (CHR) for ScZ that are currently being recruited as part of the ongoing, MRC-funded Youth Mental Health Risk and Resilience Study (YouR-Study) (Uhlhaas et al. 2017). The YouR-study aims to identify neurobiological mechanisms and predictors of psychosis-risk with a state-of-the-art neuroimaging approach (Magnetoencephalography, Magnetic Resonance Spectroscopy, Magnetic Resonance Imaging) in combination with core psychological processes, such as affect regulation and attachment, that have been implicated in the development and maintenance of severe mental health problems. In addition, genetic and metabolomic information is available from blood and urine samples.

Recruitment started in 2014 involving clinical services and investigators at Glasgow and Edinburgh Univ. (PI: P. Uhlhaas; Co-PI: S. Lawrie) and is scheduled to be completed in 2019. In addition to CHR-participants, a group of controls (n = 50) as well as a help-seeking sample of n = 40 have been recruited. Follow-up data for CHR- and help-seeking participants are available for up to three years that provide data on diagnostic outcomes, stress-levels and interpersonal functioning and affect regulation.

The proposed project will utilize and integrate information from clinical variables, neuroimaging and genetics to develop multi-model algorithms that allow 1) the stratification of CHR-participants based on clinical and neuroimaging data into sub-groups that allow insights into the underlying biological pathways conferring psychosis-risk and 2) the identification of a novel, multimodal biomarker for prediction of mental health outcomes, in particular psychosis. To this end, advanced bioinformatics and machine-learning algorithms shall be employed that identify data-driven patterns obtained from ongoing analysis of neuroimaging, clinical and genetic analysis.

Training outcomes:
This project will offer a combination of training opportunities in mental health research, neuroimaging and bioinformatics in an interdisciplinary team that includes, psychiatrists, psychologists and neuroimaging researchers at Glasgow and Edinburgh University.

1. Fusar-Poli, P. et al. (2013). The psychosis high-risk state: a comprehensive state-of-the-art review. JAMA Psychiatry, 70, 107-20.
2. Uhlhaas, P.J., Gajwani, R., Gross, J., Gumley, A.I., Lawrie, S.M., Schwannauer, M. The Youth Mental Health Risk and Resilience Study (YouR-Study). BMC Psychiatry. 26;17(1):43.

Development of a measure of treatment burden and patient capacity in stroke

Prof Frances Mair, Institute of Health and Wellbeing, University of Glasgow
Dr Katie Gallacher, Institute of Health and Wellbeing, University of Glasgow
Dr Terry Quinn, Institute of Cardiovascular and Medical Sciences, University of Glasgow
Mrs Heather Murray, Robertson Centre for Biostatistics, Institute of Health and Wellbeing, University of Glasgow

PhD Project Summary

Advances in science and technology have led to improved treatments for stroke survivors, but an increasing workload is being placed on these individuals and their carers as they are required to follow complex treatment regimes. For example, the prescribing of multiple medications following a stroke is routinely recommended to promote recovery and prevent another stroke occurring, and this can result in drug-drug interactions. Additionally, people with stroke are often discharged home early from hospital to continue therapies in the community without the support of a hospital setting. Treatment burden is defined as the workload of healthcare for people with long term conditions such as stroke and the impact of this on wellbeing [1,2]. Individuals differ in their abilities to manage their lives and health problems depending on a variety of physical, psychological, environmental and social factors; this may vary over time and is defined as patient capacity [2]. For example, those who are well supported by family and able to pay for additional support are likely to have a larger capacity that those who are social isolated and living in poverty.

Current stroke guidelines lack a person-centered approach and important factors such as treatment burden and patient capacity are not adequately considered by clinicians and policy makers. Excessive treatment burden can lead to negative outcomes such as reduced quality-of-life, non-adherence, less effective treatment and wasted resources [3].

Previous qualitative systematic review and interviews with stroke survivors examined the experience of treatment burden and the factors that affect patient capacity. Results showed that treatment burden has four components: sense making, interacting with others, enacting treatments and reflecting on progress [1,4]. Treatment burden can occur as a consequence of increased healthcare workload e.g. multiple healthcare appointments, or deficiencies in care e.g. long waiting times [1,4]. Treatment burden is influenced by patient capacity. Six factors affect patient capacity: personal skills and attributes; physical and cognitive abilities; social support; financial status; life workload and environment [2,4]. The design and delivery of health services has a considerable impact on treatment burden [1,2,3,4]. Research is needed to develop ways of measuring treatment burden and patient capacity in people with stroke that would be appropriate for use as an outcome measure in clinical trials of stroke treatments and complex interventions. Such a measure could also be used in a clinical context to screen for those experiencing high burden or low capacity.

To develop and validate a patient reported measure of treatment burden and patient capacity in people with stroke.

This can be broken down into the following objectives:
1) Systematic review of published measures of treatment burden and patient capacity in people with stroke.
2) Item generation for the new measure.
3) Cognitive interviewing to determine whether questions and instructions are relevant and coherent.
4) Validation of the measure in a sample of individuals who have had a stroke.

Training outcomes:
Under the guidance of an experienced and interdisciplinary supervisory team, the student will have an excellent opportunity to develop key generic research skills as well as expertise in patient reported measurement development. Dr Terry Quinn is coordinating editor of the Cochrane Dementia group and a member of the NIHIR Complex Review Support Unit, therefore the successful applicant would benefit from an established training program in basic and advanced systematic review. The student would also have access to The Stroke Association public engagement officer, with workshops and support available to facilitate patient, public and carer involvement in research. They will receive specialist training in patient reported measure development including cognitive interviewing and statistical analysis. Training will be provided to support the dissemination of research, for example presentation at conferences and publication in peer review journals. Students will be welcomed into a supportive and experienced research team.

1. Gallacher, K., et al., Uncovering Treatment Burden as a Key Concept for Stroke Care: A Systematic Review of Qualitative Research. PLoS Med, 2013. 10(6): p. 10.1371/journal.pmed.1001473.
2. May, C.R., et al., Rethinking the patient: using Burden of Treatment Theory to understand the changing dynamics of illness. BMC.Health Serv.Res., 2014. 14: p. 281
3. May, C., V.M. Montori, and F.S. Mair, We need minimally disruptive medicine. BMJ, 2009. 339: p. b2803.
4. Gallacher, K., An exploration of treatment burden and patient capacity in people with stroke, in Institute of Health and Wellbeing. 2016, University of Glasgow: Glasgow.

Drug induced mutational burden in cancer: understanding the risks of recurrence and drug resistance to better control them

Dr David Chang, Institute of Cancer Sciences, University of Glasgow
Prof Martin Taylor, MRC Human Genetics Unit, IGMM, University of Edinburgh
Dr Susanna Cooke, Institute of Cancer Sciences, University of Glasgow

PhD Project Summary

Genome stability must be ensured at all times to prevent carcinogenesis. Unfortunately, cells are constantly exposed to genotoxic stimuli and DNA is systematically and heavily modified or insulted at each cell cycle. While eukaryotic cells have evolved plethora of mechanisms to prevent, or recognize and repair, DNA damage, mutations in tumor suppressors such as BRCA1, BRCA2, or TP53 weakens a cell’s ability to detect DNA damage and repair it, allowing accumulation of mutations and ongoing genome instability.

Accumulation of DNA damage can lead to collapse of replication forks, gross chromosomal rearrangements, and sequence-level changes. Eventually multiple ‘hits’ are acquired in cancer genes, resulting in outgrowth of an overt malignancy. However, mutations in key genes causing defectiveness in DNA damage repair pathways, such as the homologous recombination pathway, also render tumors exquisitely sensitive to genotoxic chemotherapy and ionizing radiation, in a manner akin to synthetic lethality. In addition, such defects can synergize with inhibitors of other DNA repair pathways, for example such PARP inhibitors in BRCA1/2 mutation carriers. This molecular aberration is exploited in the treatment of hereditary cancers of the pancreas, the breast, or ovaries [1].

Most chemotherapies are highly toxic to all cells, and their efficacy as a treatment relies on the higher mitotic index of cancer cells, which sensitizes them to treatment. Importantly, chemotherapy creates breaks, adducts or mutations in the DNA of normal cells that causes treatment side effects and can lead to secondary cancers over time.

While it is critical to prevent drug-induced secondary tumors following early onset cancers (e.g., pediatric cancers, or some subsets of breast) the low survival at 5 years in pancreatic cancer has made it impossible to study the long-term effects of neoadjuvant/adjuvant therapy or new treatments alike on normal human cells. Here, we propose to revisit this important question. Having a better understanding of the currently unquantifiable consequences of drug treatment could lead us to a better understanding of the mechanism of action of existing and novel therapies, and could help delineate underlying mechanisms of treatment based on individual genetic backgrounds. Through this project, we will draw human maps of drug-induced mutations. By doing so, we will establish a pipeline to predict and possibly address the inevitable event of drug resistance [2]. This project will leverage on the University of Glasgow lead pan-UK precision medicine project for pancreatic cancer, Precision-Panc (, and it is envisaged there will be more than 1,000 patients recruited to the platform over the next 5 years who are fully genomically profiled with accompanying complete phenotypic follow up data. Majority of these patients will have post treatment genomic profile before they go onto the next phase or line of therapy, which makes it ideal to study the treatment effects and acquired resistance mechanism.

Aim 1: The supervisory team has >50 patient-derived cell lines which are fully characterized genomically, transcriptomically and epigenomically. There are also a range of pancreatic stellate cells (pancreatic cancer stromal cells) available to co-culture. We will assess the impact of chemotherapy regimens on tumor and adjacent tissues through the establishment of co-culture models in 2D and 3D. This will allow the student to test the effects of drugs and/or radiotherapy. Mutation load and spectrum read-out will be based in the first instance on both whole genome and target captured sequencing with a view to developing maximally informative targeted sequencing assays. Both the short and long-term effects of therapies in these ex vivo systems will be evaluated.
Aim 2: The supervisory team has a strong track record in computational analysis identifying molecular signatures and mutational events that can be used to classify cancer into subtypes of prognosis and therapeutic responsivenss [3,4]. Building on such expertise, the student will investigate general mutational signatures and their link with the newly established replication defective patterns identified in squamous subtype pancreatic cancer (manuscript in preparation). Emphasis will be placed on investigating large groups of genome structural rearrangements, such as transitions and transversion events, base modifications, or incorporation of uracil or amino-purines, in different subtypes of pancreatic cancers.
Aim 3: In addition to large families of mutations, we will interrogate discrete events that are often overlooked in whole genome studies. Through the effort of Precision-Panc, ~1,000 fully characterized patients with chemotherapy, outcomes and response data will be made available to interrogate rare traits, and use sub-clonal events to call signatures using bespoke panels from a smaller genomic footprint. This aim will be conducted on several cancer subtypes, in naïve tumors and correlated with response to treatment.

Training outcomes:
The trainee will benefit from being exposed to a translational pipeline. Both Glasgow- and Edinburgh- based laboratories are highly collaborative, and bring a breadth of expertise across clinical research, bioinformatics, and systems biology. By the end of their PhD, student will be fully trained in screening technologies (wet lab) as well as in high-throughput data intensive analysis for Health-related research projects.

1. Pancreatic Cancer Genomes: Implications for Clinical Management and Therapeutic Development. Dreyer SB, Chang DK, Bailey P, Biankin AV. Clin Cancer Res. 2017 Apr 1;23(7):1638-1646.
2. BRCA2 secondary mutation-mediated resistance to platinum and PARP inhibitor-based therapy in pancreatic cancer. Pishvaian MJ, Biankin AV, Bailey P, et al. Br J Cancer. 2017 Apr 11;116(8):1021-1026.
3. Genomic analyses identify molecular subtypes of pancreatic cancer. Bailey P, Chang DK, Nones K, et al. Nature. 2016 Mar 3;531(7592):47-52. doi: 10.1038/nature16965. Epub 2016 Feb 24.
4. Mutational Biases Drive Elevated Rates of Substitution at Regulatory Sites across Cancer TypesKaiser, V. B., Taylor, M. S. & Semple, C. A. 4 Aug 2016 In : PLoS Genetics. 12, 8, p. e1006207.

Genome-wide, gene-specific patterns of glucorticoid-receptor coregulators critical for growth and skeletal development

Prof S. Faisal Ahmed, School of Medicine, University of Glasgow
Dr Pawel Herzyk, Institute of Molecular, Cell and Systems Biology, University of Glasgow
Prof Colin Farquharson, Roslin Institute, R(D)SVS, University of Edinburgh
Prof Lars Savendahl, Karolinska Institutet, Stockholm, Sweden

PhD Project Summary

Glucocorticoids (GCs) are a class of steroid hormones that bind and activate the intra-cellular GC receptor (GR), and regulate transcription of many genes that govern the effects of GCs. The remodeling of chromatin and regulated assembly of active transcription complexes by GR and other transcription factors may be modulated by several hundred transcriptional co-regulator proteins and the conventional concept of transrepressors and transactivators may be too simplistic. This concept of differential regulation forms the basis of a novel group of drugs, the selective GC receptor modulators (SGRMs) that have the potential to precisely tailor the pharmacological outcome to the medical needs of the patient (1). However, there is very little information on the gene-specific patterns that may be activated following exposure to SGRMs (2). Previous work by our group has focused on understanding the effects of GCs on growth and skeletal development through a range of models including chondrogenic cell lines, ex vivo murine metatarsal cultures as well as in vivo whole animal studies in mice (3). This has also included a study of the effects of SGRMs on skeletal development (4). Going forward, the adoption of a non-selective genomic approach will allow us to develop a high-throughput approach that could improve our knowledge of SGRMs as well as GC action on growth and skeletal development.

The project relies on three-way collaboration between Glasgow, Edinburgh and Stockholm. The aim of the project will be to perform global analysis of GC-regulated gene expression in cells (osteoblasts and chondrocytes) and whole bone tissue maintained in culture. Cells and tissue will also be obtained from on-going in vivo studies in Edinburgh and Stockholm that involve mice being exposed to GC and a novel SGRM (MTA approved) at different stages of development. In addition, depletion of function of critical co-regulators will be performed using siRNAs as well as the CRISPR/Cas9 gene editing system to knock-out targeted genes in individual cells. The project will be supported through existing work that is being undertaken as part of a MRC Clinical Training Fellowship at the Roslin Institute under the joint supervision of Professors Ahmed and Farquharson. Pawel Herzyk will provide expertise and environment for RNAseq experiments, use a range of rank-based methods recently optimized for RNAseq data, and will provide bioinformatics training to the new PhD student to make him/her capable of performing integrative data analysis with various large-scale sequencing data. In summary, the PhD student will build on the existing studies to compare the genomic fingerprint that is associated with an adverse growth and bone profile following GC exposure and compare that to the fingerprint obtained following SGRM exposure.

Training outcomes:
The PhD student will join a research group with a high level of interdisciplinary skills in bone and growth plate biology, endocrinology and genomics and bioinformatics. Specifically, the student will develop project specific research skills where s/he will become proficient in the use of cell based and murine models to study GC actions. By the end of the studentship the student will have an advanced level of knowledge of bone and the growth plate and competent in cell and organ culture, real-time PCR, western blotting, immunohistochemistry and genomic and bioinformatic analysis. Ongoing formal and practical training in skills and techniques will be provided by all supervisors and members of their research groups during the studentship. The student will also develop transferable research skills through the Research Training and Personal Development Programme available through the College. The student will participate and present their data at group meetings, journal clubs and external scientific meetings.

1. Ayroldi E, Macchiarulo A, Riccardi C. Targeting glucocorticoid side effects: selective glucocorticoid receptor modulator orglucocorticoid-induced leucine zipper? A perspective. FASEB J. 2014;28:5055-70.
2. Wu DY, Ou CY, Chodankar R, Siegmund KD, Stallcup MR. Distinct, genome-wide, gene-specific selectivity patterns of four glucocorticoid receptor coregulators. Nucl Recept Signal. 2014;12:e002.
3. Wood CL , Soucek O, Wong SC, Zaman F, Farquharson C, Savendahl L, Ahmed SF. Animal models to explore the effects of glucocorticoids on skeletal development. J Endocrinol. Under revision
4. Owen HC, Miner JN, Ahmed SF, Farquharson C. The growth plate sparing effects of the selective glucocorticoid receptor modulator, AL-438. Mol Cell Endocrinol. 2007;264:164-70.

Heart Failure – The Importance of Uncertainty to Precision Medicine in Redefining Britain’s Most Common Deadly Disease

Prof John GF Cleland, Robertson Centre for Biostatistics & Clinical Trials / Institute of Health and Wellbeing, University of Glasgow
Dr Maria Wolters, Centre for Design Informatics and Institute for Language, Cognition, and Computation, School of Informatics, University of Edinburgh; Alan Turing Institute for Data Science, London
Dr Sarah Barry, Robertson Centre for Biostatistics & Clinical Trials / Institute of Health and Wellbeing, University of Glasgow
Dr David McAllister, Institute of Health and Wellbeing, University of Glasgow

PhD Project Summary

This PhD project, under the guidance of a powerful inter-disciplinary supervisory team, of clinicians and data-scientists [from The Robertson Centre for Biostatistics and Clinical Trials Unit, the Glasgow node of the Farr Institute in Scotland, the School of Informatics, Edinburgh, and the Alan Turing Institute for Data Science] will provide the student with a sound basis of data-science applied to medicine.

Precision Medicine (delivery of the right treatment to the right patient at the right time in the right amount) depends on the certainty of the diagnosis, which often reflect 19th Century definitions of disease. Lord Kelvin (1824-1907) said, “When you can measure what you are speaking about, and express it in numbers, you know something about it…”.
Heart failure is a growing challenge, not only because of changing demographics and increased longevity with chronic disease but also because it has no robust 21st Century definition that is fit for clinical or scientific purpose and yet its disorganized management consumes billions of healthcare dollars world-wide. Longitudinal studies suggest that one in every five people will develop something called “heart failure” (1), making it Britain’s most common deadly disease (2). It is probably under-recognized and diagnosed late (or not at all).

Conventionally, heart failure is diagnosed based on symptoms such as exertional breathlessness and ankle swelling (sensitive but not specific), physical signs (specific but not sensitive), echocardiograms (subject to measurement and classification errors) and biomarkers such as natriuretic peptides (a recent and evolving introduction that is not yet widely used). The diagnostic label of heart failure is usually applied late, only when symptoms and signs are severe enough to require hospitalization (3) and even then, the diagnosis is frequently missed (4). Diagnostic rates in primary care vary widely with up to a five-fold difference in prevalence amongst practices ( It is unlikely that a definition for heart failure of practical clinical use will be agreed any time soon.

About 1% of adults in the population are said to have heart failure (the disparity between incidence and prevalence is explained by high mortality) and yet 3-4% are taking high-ceiling diuretics (5), for which there is no good indication other than heart failure. Most patients who die in the years after a myocardial infarction first develop heart failure (6) and 20% of people with long-standing hypertension or diabetes will have an elevated plasma NT-proBNP (7); a powerful indicator of cardiac dysfunction.

An alternative to trying to create an artificially robust set of diagnostic criteria that may not apply to most people with the disease is to define levels of (un)certainty. Defining heart failure as a possible, probable or definite diagnosis should avoid giving false assurances and identify which patients require further diagnostic and/or therapeutic intervention. The ultimate expressions of heart failure are recurrent hospitalization for worsening symptoms and signs and death rather than the vagaries of clinical practice and expert opinion. Diagnostic patterns associated with favourable outcomes casts doubt on the diagnosis and indicates a patient group who may not benefit from intensive management. Diagnostic patterns associated with an unfavourable outcome indicate unmet need and that further investigation is required if the diagnosis is unclear.

Nationally, Scotland has a unique national prescribing information system that can be linked to both hospital and primary care information systems and mortality data. Regionally, Greater Glasgow & Clyde SafeHaven has a data-warehouse containing laboratory (haemoglobin, urea, creatinine, troponin and, more recently, natriuretic peptides) and imaging data (and stored blood) that can be linked to the clinical (eg:- hypertension, myocardial infarction, renal dysfunction) and prescribing (eg:- diuretics) records.

With expert advice from cardiologists and primary care physicians on how to classify patients, a scoring system based on routinely collected data (recognizing that key data are often missing from the patient record) will be developed that provides information on the epidemiology and outcome of possible, probable and definite heart failure, and that could be implemented in clinical practice. This will have many advantages to a ‘binary’ (yes/no) diagnosis of heart failure. It identifies people who might benefit from further investigation to change the level of diagnostic certainty and therefore therapeutic advice. Earlier diagnosis will focus greater attention on treatments to delay or prevent progression. Stratifying the diagnosis (for instance by ventricular phenotype or heart rhythm) may also allow better tailoring of treatment. Precision medicine is undermined by arbitrary and inaccurate diagnosis; quantifying diagnostic uncertainty improves precision. Creation of a grey-zone of uncertainty reduces the chance of extreme misclassification.

This proposal will provide the candidate with the opportunity to work with an alliance of clinicians, statisticians and bioinformatics focused on a burning clinical problem of global importance. The size, number and complexity of the associated data-sets will require the student to acquire new skill-sets, tools and understanding.

The completion of this studentship will equip the student with a deep knowledge of the epidemiology of heart failure and the data science skills required to describe it. The specific tasks for the student will be: 

  1. To develop a generic tool that can mine existing data-sets to anticipate and identify cases with possible, probable and definite heart failure and use these to describe the epidemiology and natural history of disease, including likely cause of death. Clinicians with expert knowledge on the epidemiology and death adjudication will help inform this exercise (eg:- diagnoses prior to and place of death are highly informative). This will be vital for large scale precision medicine and population-based genetic studies. 
  2. To develop tools useful in predicting hospitalizations and mortality. 
  3. To identify individuals of interest who have provided blood samples to the Glasgow biorepository and consent for their research use. Biomarkers, established and experimental, will be assessed to determine how they influence classification. 
  4. To employ Bayesian modelling, machine learning techniques and other inductive (data driven) methods to identify new relationships in prognostic modelling alongside traditional deductive (hypothesis driven) methods.

Training outcomes:
This PhD project, under the guidance of a powerful inter-disciplinary supervisory team, comprising both clinicians and data-scientists, from The Robertson Centre for Biostatistics and Clinical Trials Unit, the Glasgow node of the Farr Institute in Scotland, the School of Informatics, Edinburgh, and the Alan Turing Institute for Data Science will provide the student with a sound basis of data science applied to medicine. The student will specifically be trained and develop competencies in epidemiology, medical informatics, statistics and routinely collected health data. The student will also be trained in aspects of human factors and computer supported cooperative work relevant for the creation of useful tools. This will take place at the Edinburgh Center for Design Informatics, which is unique in the UK for its focus on designing with data. There will also be opportunities to develop skills in disseminating the results of analyses within the R/Shiny environment, including producing dynamic documents and web apps. This studentship will use the existing Scottish Safe Haven framework giving experience to the student of current best practice for securely handling unconsented patient clinical information. The student will obtain skills crucial for analysing large complex datasets, including version control and literate programming. The student will be encouraged to spend 3-6 months as an enrichment student at the Alan Turing Institute for Data Science.


  1. Lloyd-Jones DM, Larson MG, Leip MS, Beiser A, D'Agostino RB, Kannel WB, Murabito JM, Vasan RS, Benjamin EJ, Levy D. Lifetime Risk for Developing Congestive Heart Failure - The Framingham Heart Study. Circulation 2002;106(24):3068-3072.
  2. Cleland JGF, McDonagh T, Rigby AS, Yassin A, Whittaker T, Dargie HJ, on behalf of the National Heart Failure Audit Team for England and Wales. The National Heart Failure Audit for England and Wales 2008-2009. Heart 2011;97(11):876-886.
  3. Cowie MR, Wood DA, Coats AJS, Thompson SG, Poole-Wilson PA, Suresh V, Sutton GC. Incidence and aetiology of heart failure. A population based study. Eur Heart J 1999;20:421-428.
  4. Khand AU, Shaw M, Gemmel I, Cleland JG. Do discharge codes underestimate hospitalisation due to heart failure? Validation study of hospital discharge coding for heart failure. Eur J Heart Fail 2005;7(5):792-797.
  5. Clarke KW, Gray D, Hampton JR. How common is heart failure? Evidence from PACT (Prescribing Analysis and Cost) data in Nottingham. Journal of Public Health Medicine 1996;17(4):459-464.
  6. Torabi A, Rigby AS, Cleland JGF. Declining In-Hospital Mortality and Increasing Heart Failure Incidence in Elderly Patients with First Myocardial Infarction. J Am Coll Cardiol 2009;55(1):79-81.
  7. McGrady M, Reid CM, Shiel L, Wolfe R, Boffa U, Liew D, Campbell DJ, Prior D, Stewart S, Krum H. NT-proB natriuretic peptide, risk factors and asymptomatic left ventricular dysfunction: results of the SCReening Evaluation of the Evolution of New Heart Failure study (SCREEN-HF). Int J Cardiol 2013;169(2):133-138.

How does a trypanosome change its spots? Decrypting immune avoidance in human trypanosomes

Dr Martin Llewellyn, Institute of Biodiversity, Animal Health and Comparative Medicine, University of Glasgow
Dr Richard McCulloch, Institute of Infection, Immunity & Inflammation, University of Glasgow
Dr Richard Burchmore, Institute of Infection, Immunity & Inflammation, University of Glasgow
Dr Caroline Gauchotte-Lindsay, School of Engineering, University of Glasgow
Dr Bjorn Andersson, Karolinksa Institut

PhD Project Summary

Avoidance of host immunity is crucial for trypanosome survival. Trypanosoma brucei possesses elaborate and well-characterised strategies to achieve this. The mechanism of immune avoidance in Trypanosoma cruzi is largely unknown: arcane gene expression and high genomic complexity frustrate efforts to reconstruct surface antigen diversity. For both T. brucei and T. cruzi, our existing understanding of surface protein diversity is based to a significant extent on genomic and transcriptomic approaches. Targeted methodologies to directly characterize the surface proteome are lacking. Such methodologies are potentially transformative for understanding phenomena fundamental to parasite biology and control e.g. immune evasion, host cell entry, passage through the lifecycle.

Due to their insolubility in aqueous solutions and natural low abundance, membrane glycoproteins (MGPs) are extremely difficult to separate or analyse by traditional methods. Given adequate solubilisation of membrane fractions, micro-scale capillary electrophesis (CE) can provide significant insights into the composition and abundance of MP classes in a given low abundance sample. To improve throughput and sensitivity, reduce reagent costs, improve separation efficiency and reduce analysis time, protein CE is scaled down to operate on a chip – so called micro-CE. Such devices can be adapted to simultaneously analyse multiple sub-fractions from the same membrane sample, providing unprecedented detail into membrane protein composition.

In this PhD project the student will use a high-throughput microfluidic chromatography approach to characterise the surface glycoproteome of T. cruzi, T. brucei and Lesihmania cells exposed to host immunity. Changes in the surface glycoproteome will correlated with changes in transcriptomic diversity in the same cell lines to establish to true impact of immune evasion on cells surface antigenic reprtoires.

Training outcomes:
The student will receive training in genomics and transcriptomics with Drs Llewellyn, McCulloch in the UK and Prof. Andersson in Sweden, proteomics with Dr. Burchmore, and microfluidics design with Dr. Gauchotte-Lindsay. The training opportunities in this project will provide the student with a diverse ‘omics skillset, as well as skills in detector development and deployment and should form an excellent foundation for a future career in biomedical science.

Identifying Inflammatory Bowel Disease Pathotypes

Prof Simon Milling, Institute for Infection, Immunity and Inflammation, University of Glasgow
Prof Rob Nibbs, Institute for Infection, Immunity and Inflammation, University of Glasgow
Dr Stefan Siebert, Institute for Infection, Immunity and Inflammation, University of Glasgow
Prof Daniel Gaya, NHS GG&C, Associate Clinical Professor

PhD Project Summary

Inflammatory bowel disease is driven by T lymphocytes and cells of the innate immune system, particularly macrophages. Two separate clinical conditions are recognised; ulcerative colitis (UC) affects only the colon and rectum, while Crohn’s disease (CD) can affect any region along the intestine. Both conditions are common, cause major morbidity and are rapidly increasing in prevalence in Westernised countries. Recently, immunotherapeutic approaches have shown promise in the treatment of CD and UC1. Licensed antibodies block either soluble inflammatory molecules (anti-TNF, anti-IL12/23), or inhibit lymphocyte migration to the intestine (anti-47 integrin). Unfortunately, these therapies are expensive, and induce remission in only 1/3 of treated individuals. New classes of antibodies and small molecules2 are in development, but these also only appear to be effective for a similar proportion of patients. There is little understanding of why many patients do not respond. Thus, there is a huge unmet need for novel strategies to stratify patients, and to generate better treatments.

Significant progress has been made in understanding the mechanisms of intestinal inflammation3. This can be exploited to understand the pathological pathways that drive CD and UC. We have recently developed the following resources, which will contribute to this project:

  • In the mouse, we have identified the specific lymph nodes from which T cells migrate to the colon and the small intestine3. We have generated RNAseq datasets from these T cells, to identify molecules directing T cell homing to the colon. This mouse work is central to our on-going MRC project (see above).
  • From Gastroenterologists led by Daniel Gaya we obtain weekly colonic biopsies from healthy individuals, and biopsies and blood samples from CD and UC patients, From the GG&C Biorepository we receive samples from bowel resections. All samples are accompanied by comprehensive clinical data, with permission for follow-up enquiries.
  • We have characterised purified macrophage populations from human tissue samples. RNAseq datasets have been generated from two different human macrophage populations. 
  • We have generated a comprehensive immunophenotype using blood from CD patients, compared to heathy individuals and other inflammatory diseases. This has identified specific T lymphocyte populations as important for pathology in CD.
  • We have a close collaboration with Boehringer Ingelheim (BI), based in Ridgefield (CT, USA), whereby we have received RNAseq data from some healthy, CD and UC biopsies. BI are keen to support a PhD student who will contribute to extending these studies.

We will use these resources to address the following aims:

1) To compare the characteristics of biopsies from CD and UC patients with their subsequent responses to treatment. Cellular phenotypes and gene expression data will be generated from intestinal biopsies from CD and UC patients. Data will be correlated against treatment and response data, collected at the time of biopsy and at subsequent visits.
2) To test whether colon-specific molecules expressed by T cells will be expressed in colonic T cells from patients with CD and UC, and whether these may be useful for stratification or therapeutic targeting.
3) To identify markers, in repeat blood samples collected before and after treatment, that correlate with disease pathotypes identified in the biopsy samples above, or with response to treatment, by blood immuno-phenotyping, gene expression analysis, and using multiplex arrays.

Training outcomes:
The student will receive training at the technical level (tissue sampling and preparation, flow cytometry, gene expression analysis, multiplex technology, analysis of bio-informatic data), the clinical level (research ethics, analysis of anonymised patient data, hypothesis testing), in industrial science (understanding the therapeutic pipeline, industrial decision-making, ‘big’ science) and in more generic skills (interactions with patient, nurses, clinicians and scientists, presentation of results, writing).

1. Neurath MF. Current and emerging therapeutic targets for IBD. Nat Rev Gastroenterol Hepatol. 2017 May;14(5):269-278.
2. Patterson H, Nibbs R, McInnes I, Siebert S. Protein kinase inhibitors in the treatment of inflammatory and autoimmune diseases. Clin Exp Immunol. 2014. Apr;176(1):1-10.
3. Houston SA, Cerovic V, Thomson C, Brewer J, Mowat AM, Milling S. The lymph nodes draining the small intestine and colon are anatomically separate and immunologically distinct. Mucosal Immunol. 2016 Mar;9(2):468-78.

Investigating and addressing non-representativeness of Understanding Society longitudinal study data via multi-source record linkage

Dr Linsay Gray, Institute of Health and Wellbeing, University of Glasgow
Dr Frank Popham, Institute of Health and Wellbeing, University of Glasgow

PhD Project Summary

Panel studies - which follow and gather in-depth information on the same people over time - provide rich longitudinal data for research. These data are critical to measuring population metrics such as health, social factors and other related determinants for informing social, economic and health policy and practice. However, incomplete participation at recruitment and non-response over time threaten validity of estimates (Christensen, Ekholm et al. 2015), with prime targets for precision medicine – such as high disease risk sub-groups – likely to be disproportionately under-represented. We have recently developed a novel tailored response bias-correction methodology which utilises record-linkage and multiple imputation (Gorman, Leyland et al. 2014, Gorman, Leyland et al. 2017). The methodology has previously been applied to cross-sectional survey data and this PhD would extend its development and application for use in longitudinal studies, with additional use of inverse probability weighting. Understanding Society is an innovative world-leading panel study – the largest of its kind – providing valuable information about 21st century UK life over time (Buck 2008). The study follows the lives of individuals from 40,000 UK households, collecting data covering multidimensional aspects of the lives of individuals and their families including measures of physical and mental health, biomarkers, genetics and epigenetics as well as demographic and socio-economic characteristics. Individual level record-linkage of these data to other sources, including administrative health and education data, has been initiated. The PhD programme will involve adaptation and application of the methodology for addressing non-participation across six waves of Understanding Society. This will lead to the obtaining of more reliable estimates of health and clinical measurements. There is potential for wider application to other longitudinal study data in this era of declining participation.

This PhD programme aims to address non-representativeness in Understanding Society data to resolve bias arising from non-participation and non-response by way of the customisation of dedicated statistical methodology, exploiting the record-linked resource.

The studentship will:
1) Quantify the differences in health outcomes (mortality, hospital admissions) between Understanding Society participants and the general population;
2 Integrate insights gained to correct for non-representativeness in Understanding Society data using multiple imputation and inverse probability weighting techniques; and
3) Generate study-based estimates of health and clinical indicators corrected for non-representativeness.

Training outcomes:
This programme will provide the basis for training in advanced statistical methods including handling missing data and bias which although commonplace in observational data-based research are under-skilled areas. The student will also acquire knowledge of the determinants of health and social inequalities in health. This will enable them to develop the expertise required to undertake the project, to understand how their work fits in within the broader perspective of health inequalities, and help develop them as a future independent scientist. The supervision team brings together researchers with expertise in statistics (Gray and Popham), population health sciences (Gray and Popham) and social sciences and social policy (Popham). Gray has particular expertise in the development and application of methodology for advancing understanding of the determinants of health and improving population health and inequalities, with specific interest in analysis of record-linked survey and administrative data. This project will also build on an established collaboration with colleagues at the Institute for Social and Economic Research, University of Essex (chiefly Director of Understanding Society, Professor Michaela Benzeval).

1. Buck, N. (2008). Understanding Society: design overview. Understanding Society Working Paper Series, University of Essex. 2008 – 01.
2. Christensen, A. I., O. Ekholm, L. Gray, C. Glümer and K. Juel (2015). "What is wrong with non-respondents? Alcohol-, drug- and smoking-related mortality and morbidity in a 12-year follow-up study of respondents and non-respondents in the Danish Health and Morbidity Survey." Addiction 110(9): 1505-1512.
3. Gorman, E., A. H. Leyland, G. McCartney, S. V. Katikireddi, L. Rutherford, L. Graham, M. Robinson and L. Gray (2017). "Adjustment for survey non-representativeness using record-linkage: refined estimates of alcohol consumption by deprivation in Scotland." Addiction [Epub ahead of print].
4. Gorman, E., A. H. Leyland, G. McCartney, I. R. White, S. V. Katikireddi, L. Rutherford, L. Graham and L. Gray (2014). "Assessing the representativeness of population-sampled health surveys through linkage to administrative data on alcohol-related outcomes." American Journal of Epidemiology 180(9): 941-948.


Role of the alternative splicing in the gametocyte production in malaria parasites

Dr Katarzyna Modrzynska, Institute of Infection, Immunity & Inflammation, University of Glasgow
Prof Andy Waters, Institute of Infection, Immunity & Inflammation, University of Glasgow
Dr Pawel Herzyk, Institute of Molecular, Cell and Systems Biology, University of Glasgow

PhD Project Summary

Mosquito-transmitted unicellular parasites from the Plasmodium genus are best known as causative agents of malaria, one of the most deadly diseases in tropics. They are characterised by the complex life cycle, involving multiple life stages of varying size, shape and motility, and a compact genome with ~ 5000 genes. Interestingly more than half of these genes (~55%) contain introns, often multiple ones, raising the possibility that the parasite uses alternative splicing to expand its transcript pool. Indeed, for a number of genes, multiple splicing isoforms were observed in recently published transcriptomes of both human and rodent malaria parasites [1], [2]. Their significance in producing different life stages, however, was not investigated so far.

Of particular interest is the role alternative splicing may play during the Plasmodium development in the mammalian blood. There, the genetically identical parasites can transform into either asexual forms, contributing to the disease symptoms, or into male or female gametocytes responsible for the transmission through the mosquito. While the key switch between the asexual and sexual pathway was identified recently [3], the mechanisms of the sex determination remain one of the biggest unanswered questions in Plasmodium biology, as the parasite has no sex chromosomes and remains haploid through most of its development. Recently we generated a time–course of transcriptome changes during Plasmodium berghei differentiation into either asexual forms or gametocytes as well as the transcriptomes of purified mature asexual, male and female parasites. The initial analysis of these datasets revealed a number of transcripts edited differently in different cell types (especially between males and females) suggesting, that alternative splicing might play a key role in the Plasmodium gametocytogenesis (KM and APW, unpublished data).

The goal of the proposed project is to identify the alternative splicing events that take place during early gametocytogenesis and assess their importance for parasite transmission.

The student will generate a new RNA-seq dataset from the early-stage purified Plasmodium gametocytes using long-read Oxford Nanopore technology, which allows to sequence full-length transcript isoforms. The results will be combined it with the existing P.berghei and P.falciparum transcriptome datasets in order to obtain a full inventory of species-transcending splicing isoforms involved in early male/female differentiation.

The key findings from the in silico analysis will be validated using experimental rodent malaria model Plasmodium berghei which allows the easy access to all stages of the life cycle in the laboratory conditions and is very amenable to the genetic modification. DiCre- and Cas9-based genome editing systems will be used to manipulate the parasite’s gene structure and a range of phenotyping assays will be employed in order investigate the changes in the life cycle progression.

Training outcomes:
The student will have an opportunity to acquire a range of the bioinformatics skills working with diverse NGS datasets and comparing various tools for read mapping, alternative splicing analysis and data visualisation. At the same time, he or she will also become familiar with different stages of the life cycle of malaria parasites, the strategies of generation of P.berghei mutants, state of the art microscopy and flow cytometry approaches used for parasite phenotyping and NGS library preparation for different platforms.

In addition to the laboratory training, the student will be provided with opportunities to present their work both locally and internationally, develop collaboration with other groups from the WTCMP and publish in leading high impact journals. He/She will also be encouraged to make their work accessible to the general public and participate in international scientific conferences providing opportunities for networking and career development.

[1] K. Sorber, M. T. Dimon, and J. L. DeRisi, “RNA-Seq analysis of splicing in Plasmodium falciparum uncovers new splice junctions, alternative splicing and splicing of antisense transcripts.,” Nucleic Acids Res., vol. 39, no. 9, pp. 3820–35, May 2011.
[2] T. D. Otto et al., “A comprehensive evaluation of rodent malaria parasite genomes and gene expression.,” BMC Biol., vol. 12, p. 86, Jan. 2014.
[3] A. Sinha et al., “A cascade of DNA-binding proteins for sexual commitment and development in Plasmodium.,” Nature, vol. 507, no. 7491, pp. 253–7, Mar. 2014.



Routinely collected sexual health data in Scotland: integrity, data linkage and use in intervention development

Prof Lisa McDaid, Institute of Health & Wellbeing, University of Glasgow
Dr Linsay Gray, Institute of Health & Wellbeing, University of Glasgow
Associate Professor Richard Bränström, Department of Clinical Neuroscience, Karolinska Institutet
Prof Claudia Estcourt, School of Health & Life Sciences, Glasgow Caledonian University, NHS Greater Glasgow & Clyde Sandyford Sexual Health Services

PhD Project Summary

Gay, bisexual and other men who have sex with men (henceforth MSM) experience a disproportionate burden of ill health in relation to sexual, mental and physical health (1). Recent UK research demonstrated that 8.4% reported ill health in all three domains, compared with 1.5% of men reporting sex with women exclusively (2). Capturing accurate and sufficient clinical, behavioural and lifestyle data concerning the health of MSM remains a key concern and establishing how to identify those most at risk and likely to benefit from targeted biobehavioural interventions is a key aim of this studentship.

The potential for use of electronic patient records is widely recognised in the UK and there has been substantial investment in health informatics research. A centralised web-based patient management and clinical record system is in use across all specialist sexual health settings in Scotland (Scottish National Sexual Health System - NaSH). The range of clinical and lifestyle data recorded, including medical, family and lifetime and recent sexual history, reproductive health and contraception, social risk and lifestyle factors, test requests/results, patient actions/recalls, prescriptions, and partner notification, provides the scope and coverage consistent with the precision medicine approach to disease treatment and prevention. Anonymised data views of NaSH have been constructed for management reporting in NHS Boards, but our review of this data set for sexual health research highlighted issues such as: inconsistency of data collection, referential dis-integrity and instances of over-writing of single-record data with the most recently recorded value (3). Using routinely collected NaSH health records has the potential to answer valuable research questions, respond quickly to immediate and emerging health challenges, and give access to high-risk and/or hard-to-reach populations (often indisposed to research involvement), but data quality is essential and further analysis is required to assess its integrity, as well as explore solutions for reliable linkage to other data sources.

Sandyford Sexual Health Services provide all specialist sexual health care for NHS Greater Glasgow and Clyde (a NHS Board serving over one million people) and serves an estimated population of 30,000 MSM within Glasgow and the surrounding area. Over 10 years of data have been recorded, and NaSH allows patient-centred choice of whether to use an anonymous identifier or the Community Health Index (CHI) number (Scotland’s equivalent of the NHS number), which could facilitate record linkage. The data would be strengthened by linkage at the individual level to other sources such as Scottish Birth Records, Scottish Morbidity Record 01 (covering hospital admissions, caner registrations and mortality records), records of prescriptions, mental health service use, and education records, as well as measures of area-based deprivation. The studentship will formally explore the possible technical solutions, consent and practicalities of linkage of the sexual health data to other data sets.

We propose a primarily methodological studentship using data from electronic sexual health records of MSM attending specialist sexual health services in NHS Greater Glasgow and Clyde.

The studentship will:
1) explore the integrity of the data for analysis and potential linkage;
2) explore possibilities and gain appropriate consent (with support) for linkage to a range of other data sources via a range of means; and
3) measure health outcomes from longitudinal analysis (e.g. repeat attendance, disclosure of risks, diversity of sexual repertoire etc) from within the NaSH data set.

Training outcomes:
The supervisory team brings together expertise in research on measuring health inequalities (McDaid, Gray, Bränström), sexual minorities (McDaid, Bränström), sexual health (McDaid, Estcourt) and health record linkage and analysis (Gray, Bränström). Under the guidance of this experienced team, the student will receive training and develop competencies in medical informatics, routinely collected health data, record linkage and statistical analysis.

1. Public Health England. Promoting the health and wellbeing of gay, bisexual and other men who have sex with men. London: Public Health England. 2014.
2. Mercer C., Prah P., Field N. et al. The health and wellbeing of men who have sex with men (MSM) in Britain: Evidence from the third National Survey of Sexual Attitudes and Lifestyles (Natsal-3). BMC Public Health (2016) 16:525.
3. McDaid L., Docherty S., Winter A. Review of the National Sexual Health IT System (NaSH) in Scotland: the potential for sexual health research MRC/CSO Social and Public Health Sciences Unit Occasional Paper 2013: Glasgow.

Sexually dimorphic development of the human fetal brain: critical role in the programming of neurodevelopmental disorders

Dr Michelle Bellingham, Institute of Biodiversity, Animal Health and Comparative Medicine, University of Glasgow
Dr Amanda Drake, Queen’s Medical Research Institute, University of Edinburgh
Prof Paul A. Fowler, Institute of Medical Sciences, University of Aberdeen

PhD Project Summary

The fetal environment plays a key role in programming normal brain development in utero. Adverse maternal lifestyle factors such as malnutrition, obesity, smoking and alcohol consumption are associated, robustly in some cases, with neurodevelopmental and behavioural disorders such as autistic spectrum disorder (ASD) and attention deficit hyperactivity disorder (ADHD) [1]. Many psychiatric disorders have sex-specific differences in prevalence, age of onset and severity: major depressive disorder and anorexia nervosa are more common in females whilst autism spectrum disorder and alcohol dependence are commoner in males [2]. Significant differences between the male and female brain arise early in life and major sex-specific hormonal differences impact both early brain development and later brain function and behaviour. Levels of fetal brain steroids and steroid metabolising enzymes differ by developmental stage and brain region. Work in rodents has shown that altered sex steroid action during critical developmental periods impact on sexually dimorphic brain development, resulting in long-lasting effects on sex-specific patterns of behaviour [3]. Little is known, however, about normal sexually dimorphic brain development in the human and how it is regulated.

The mechanisms by which environmental factors contribute to neurodevelopmental disorders (NDs) are unclear, although epigenetic modification of DNA is likely to be involved. Maternal factors during gestation can alter epigenetic marks, changing the expression of important genes in fetal development. For example, maternal smoking is associated with altered DNA methylation (DNAm) of genes involved in placental and fetal development, leading to neurodevelopmental disorders in the offspring. Crucially, there are marked sex differences in DNAm in the brain which are established early in life as a consequence of sex steroid action. Alterations in sex steroids can disrupt the establishment of these normal sex-specific patterns of DNAm and associate with altered behaviour [4].

To understand how maternal lifestyle factors affect fetal brain development, we first need to understand normal sexually dimorphic brain development in the human. The proposed studentship will be the first to characterise patterns of normal sexually dimorphic fetal human brain development across gestation and how these are altered by maternal lifestyle factors, including smoking, obesity and deprivation status. This project will provide critical insight into the mechanisms predisposing males and females to certain behaviours and neuropsychiatric disorders in later life.

1) To characterise normal patterns of gene expression, local sex steroid concentrations and neuronal densities in male and female human fetal brain PFC across gestation (8-20 weeks).
2) To identify sex differences in DNAm profiles of candidate genes (identified in aim 1) during normal fetal development of the sexually dimorphic PFC and how these associate with sex steroid concentrations.
3) To determine how maternal smoking alters the normal patterns of sexually dimorphic fetal PFC development (as characterised in aims 1 and 2).

The proposed study will shed light on the fundamental processes underlying sex differences in human brain development and how these are altered by maternal lifestyle with longer-term impacts for drug use and development, and for public health advice.

Training outcomes:
This integrative and highly collaborative project benefits from the use of a unique, well established human fetal tissue resource which already contains a significant number of fetal brains. The student will benefit from training in the field of developmental programming and mechanisms of human brain development working with an unrivalled human tissue collection to understand the normal mechanisms of sexually dimorphic brain development, the epigenetic mechanisms of brain development and how these are affected by adverse maternal lifestyle factors such as smoking which may underlie the programming of neurodevelopmental disorders in males and females. The student will gain expertise in in-depth molecular analysis, brain imaging, epigenetic analyses, proteomics, bioinformatics, large-scale data analysis.

1. Linnet, K.M. et al (2003) Am. J. Psychiatry, 160, 1028–1040.
2. Motlagh, M.G. et al (2011) J. Atten. Disord. 15, 593–603.
3. Gobinath, A.R. et al (2017) J Neurosci Res 95, 50-64.
4. Nugent, B.M. et al (2015) Nat Neurosci 18, 690-697

The health of looked after children in Scotland

Professor Alastair H Leyland, MRC/CSO SPHSU, Institute of Health and Wellbeing, University of Glasgow 
Dr Denise Brown, MRC/CSO SPHSU, Institute of Health and Wellbeing, University of Glasgow 
Dr Mirjam Allik, Urban Big Data Centre, University of Glasgow
Ms Ruth Dundas, MRC/CSO SPHSU, Institute of Health and Wellbeing, University of Glasgow 

PhD Project Summary

Looked after children (LAC) are those children who are in care of the local authority for a period of more than 24 hours. LAC may be looked after at home or away from home (i.e. with foster or kinship carers, prospective adopters, in residential care homes, or in a residential schools or secure units). Children LA are at risk of poorer educational outcomes, greater social difficulties and reduced life chances.

Since 2000, the rate of children LA has increased substantially in Scotland; in 2015 over 15,000 (1.5%) children in Scotland were LA compared to 11,300 (0.9%) in 2000. Glasgow City has the highest rate of LAC (3.1%) with rates in Renfrewshire and West Dunbartonshire also high (2%). These areas also tend to be ones which have generally high levels of deprivation. Evidence from other countries has shown that children LA have poorer mental health and higher rates of avoidable mortality (Hjern et al., 2004, Kalland et al., 2001, Katz et al., 2011) than children not LA. Little is known about the relationship between health and being LA in Scotland, although recent work in Scotland has looked at inequalities in dental health and access to dental health services for LAC (McMahon et al., 2017).

This project will examine individual level linkage of the Scottish Government’s CLAS (Children Looked After Survey) return with health data in order to compare health-related outcomes of school-age LAC and school-age children who have never been LA between the start of the study period (2009/10) and the end of follow-up (2015/16). Children will be aged 4-17 at the beginning of the study in 2009/10 and aged 10-23 by the end of follow up in 2015/16. Both cohorts’ health outcomes will be examined in the 6-year period between 2009/10 and the end of the study. The looked after status will also be followed over the same time period from 2009/10 to 2015/16.

Data linkage for this project has been approved and is currently underway. The CLAS return is being linked to a range of health data: SMR01, 02, 04, Prescribing Information System (PIS) held by National Services Scotland, and Birth Registrations held by NRS. Hospital admissions and prescription data will be broadly grouped such as into mental health outcomes and non-mental health outcomes. The main explanatory variables of interest will come from the CLAS return for LAC (type of placement, length of placement or number of episodes etc.). Additional control variables will be available via the Pupil Census for all children (Data zone of school, school type, disability etc.) and NRS birth records (e.g. parental social class).

Health outcomes in the cohort of school-age LAC will be compared to the whole population of school-age children non-LA, while taking into account a range of important factors. Links to the Centre for Excellence for Looked After Children in Scotland (CELSIS) will help the findings from this work to be disseminated appropriately. This work is particularly timely because of the recent launch of the independent care review of Scotland’s care system.

1. Hjern, A., Vinnerljung, B. and Lindblad, F. (2003) “Avoidable mortality among child welfare recipients and intercountry adopters” Journal of Epidemiology and Community Health, 58: 412-417
2. Kalland M., Pensola, T.H., Merilainen, J. and Sinkkonen, J. (2001) “Mortality in children registered in the Finnish child welfare registry: population based study” BMJ, 323: 207-208
3. Katz, L.Y., Au, W., Singal D. et al (2011) “Suicide and suicide attempts in children and adolescents in the child welfare system” Canadian Medical Association’s Journal, 183: 1977-1981
4. McMahon A., Elliott L., Macpherson L., Sharpe K., et al., (2017) “Looked after children: inequalities in dental health needs and access to dental services: a population data-linkage study in            Scotland” Archives of Disease in Childhood, to appear

How does the health of school-age looked after children differ from the health of school-age children in the general population of Scotland?
For which health measures are differences observed?
Are there any differences by age, sex and/or deprivation?

Training outcomes:
Statistical analysis of health and social data.
Experience in the management of large linked data files.
Increased experienced in the use of R/SAS statistical software.
Safe researcher training.
Training in research ethics and governance.
Improved verbal and written communication skills.


Understanding negative chemotaxis using neutrophils & mathematics

Prof Robert Insall, Institute of Cancer Sciences, University of Glasgow
Dr Sonja Vermeren, MRC Centre for Inflammation Research, University of Edinburgh

PhD Project Summary

Negative chemotaxis is important in a range of different biomedical contexts, most importantly in the resolution of inflammation where it is usually called "reverse migration" (Nourshargh et al., 2016). When a threat has been neutralized by the immune system, it is important that leukocytes leave the area efficiently. However the signals to leave derive from the cells in the area, so must act as repellents.

It is difficult to identify mechanisms by which G-protein coupled receptors, through which chemokines and cytokines regulate chemotaxis, because unlike bacterial attractant receptors they can only convey a positive signal. Active receptors release free G-protein subunits, inactive ones don't - there is no negative option. We have recently used finite-element, PDE-based computational models (see Tweedy et al., 2016 & McDonald et al., 2016) to explore mechanisms by which negative chemotaxis could operate. Two work well - secretion of a receptor antagonist in the presence of uniformly saturating agonist levels, and, more counterintuitively, metabolising a secreted attractant that is present in near-saturating levels into a product that antagonises the same receptor.

The student will perform a detailed exploration of the mechanisms behind negative chemotaxis and reverse migration of neutrophils, using computational modelling (Insall lab) and bespoke chemotaxis assays on peripheral donor blood-derived neutrophils (an extension of an existing collaboration between the Insall and Vermeren labs, in which we have tested whether human neutrophils can solve mazes using chemoattractants) and mouse neutrophils (in which it is more straightforward to measure the effects of activation; see Chu et al., 2016). The student will:

  • Construct detailed new models in which neutrophils respond to chemokines, using parameters measured from real cells where possible. Our initial starting-point will use parameters from IL-8 chemotaxis, but we will consider others;
  • Determine which of the two basic scenarios defined above gives the most robust and broad negative chemotaxis, and seek out additional mechanisms that could recreate observed responses;
  • use the results of the models to construct real assays showing negative chemotaxis, using donated neutrophils observed by light microscopy, in devices whose dimensions are specified by the model results and the concentrations of attractants that work best;
  • compare the behaviour of unprimed and fully-activated neutrophils using untreated and thioglycollate-treated mice with induced peritonitis, to distinguish pro-inflammatory and pro-resolution processes;
  • if the ideal models involve chemokine metabolism, identify the enzymes responsible and measure their kinetic parameters.

1. G MacDonald, JA Mackenzie, M Nolan, RH Insall. 2016. A computational method for the coupled solution of reaction–diffusion equations on evolving domains and manifolds: Application to a model of cell migration and chemotaxis. Journal of computational physics 309, 207-226
2. Nourshargh S, Renshaw SA, Imhof BA. 2016. Reverse Migration of Neutrophils: Where, When, How, and Why? Trends Immunol. 37(5):273-86. doi: 10.1016/
3. L Tweedy, DA Knecht, GM Mackay, RH Insall. 2016. Self-generated chemoattractant gradients: attractant depletion extends the range and robustness of chemotaxis. PLoS biology 14 (3), e1002404.
4. Chu JY, Dransfield I, Rossi AG, Vermeren S. 2016. Non-canonical PI3K-Cdc42-Pak-Mek-Erk Signaling Promotes Immune-Complex-Induced Apoptosis in Human Neutrophils. Cell Rep. 17(2):374-386. doi: 10.1016/j.celrep.2016.09.006.

Understanding the leukaemic niche: computational modeling of intercellular signaling networks in the bone marrow microenvironment

Dr Lisa Hopcroft, Institute of Cancer Sciences, University of Glasgow
Dr Guido Sanguinetti, School of Informatics, University of Edinburgh
Dr Simon Rogers, School of Computing Science, University of Glasgow
Dr Helen Wheadon, Institute of Cancer Sciences, University of Glasgow

PhD Project Summary

Chronic myeloid leukaemia (CML) is caused by oncogene expressing (BCR-ABL+) haemopoietic stem cells (HSC) arising in the bone marrow (BM). The BM niche is a very complex ecosystem of cells, evolutionarily optimized to maintain normal function and homeostasis, and is in part maintained by intercellular communication between cell surface receptors and their corresponding, “cognate” ligands.

The BCR-ABL+ cells disrupt this signaling network to cultivate the supportive, pro-leukaemic niche necessary to maintain the disease (1). How this occurs is poorly understood but multiple relevant single cell RNA-seq (scRNA-seq) datasets (both normal and leukaemic) are now available and provide the resolution required to tackle this question. The overarching objective of this exciting interdisciplinary project is to use machine learning approaches to build probabilistic signaling networks representing intercellular communication in the BM, consider how these networks are misregulated in CML, and ultimately identify novel therapeutic interventions for the clinic.

This project will run alongside Leuka/KKLF fellowship funding awarded to Hopcroft (Institute of Cancer Sciences (ICS), University of Glasgow (UofG)) to identify targetable pathways in the CML pro-leukaemic BM niche using systems biology, and will draw on the machine learning experience of Sanguinetti (School of Informatics, University of Edinburgh) and Rogers (School of Computing Science, UofG). Sanguinetti in particular has a track record in scRNA-seq analysis (2). Wheadon (ICS, UofG) is an expert in intercellular signaling pathways in both normal and leukaemic tissues and will provide expertise and training in in vitro validation and translational outcomes.

Aim 1: To develop, optimize and test methodology to construct probabilistic intercellular signaling networks using scRNA-seq data. Using established intercellular signaling pathways, the student will develop methodologies to determine “signal flow” between/within specific populations of cells. Specifically, in the first instance, the student will design and implement a probabilistic extension of an existing bulk RNA-seq method (3) to (i) handle the unique noise/signal dropout rates of scRNA-seq data; and (ii) identify cellular subpopulations potentially active in specific intercellular communication pathways (simulated scRNA-seq data spiked in silico with specific ligand/receptor interactions will be used at first). The student will assess performance using a publicly available scRNA-seq dataset of >1500 dendritic cells, assayed both in isolation and as part of a wider cellular population (i.e., in the presence and absence of intercellular interactions) (GEO:GSE48968).
Aim 2: To construct signaling networks representing intercellular interactions between relevant cells in normal BM. Apply the methodology developed in Aim 1 to scRNA-seq data from normal, murine BM to construct a hypothesized “baseline” signaling network for the maintenance of normal haemopoiesis in murine BM. Specifically, analyse scRNA-seq data from (i) 12 haemopoietic, myeloid cell subtypes of varying maturity (GEO:GSE81682); (ii) multiple myeloid/lymphoid populations (GEO:GSE76983); and stromal cells (provided by collaborator Dr Rob Welner, University of Alabama, USA). The candidate will discuss, design and carry out appropriate in vitro experiments to validate in silico findings.
Aim 3: To assess how the signaling network is perturbed in the presence of BCR-ABL+ HSC. The Mead laboratory has assayed BCR-ABL+ HSC and BCR-ABL- HSC from CML patient BM, and HSC from healthy controls using scRNA-seq (4). The candidate will assess to what extent the signaling interactions predicted by the “baseline” network (Aim 2) are present in these primary human data to identify gain/loss of function signaling interactions that might comprise a pro-leukaemic niche. The scRNA-seq data will be stratified using clinical phenotype (specifically, response to treatment) and, again, phenotype-specific gain/loss of function signaling interactions will be identified. Appropriate in vitro validation experiments will be designed in collaboration with Wheadon and carried out by the student to verify these findings.

Training outcomes:
In this interdisciplinary project, the candidate will be fully trained in the computational methodologies necessary to generate and analyse complex ‘omics’ datasets at the bulk cell/single cell level and the downstream, in vitro validation pipeline. This will enable the DTP candidate to plan computational experimentation in the context of the translational pipeline (e.g.: Which targets should be prioritized? What kind of “dry” hypotheses can be tested in the “wet” laboratory?). It is anticipated that most of the training will be dry lab, under the supervision of Hopcroft, Sanguinetti and Rogers; this will include the full scRNA-seq pipeline (from alignment to expression quantification/analysis and downstream in silico functional analysis) and probabilistic modeling (parameter estimation, performance optimization etc). Training in in vitro experimental design and wet laboratory validation techniques (including single-cell specific assays) will be provided by Wheadon.

1. Holyoake TL, Vetrie D. The chronic myeloid leukemia stem cell: stemming the tide of persistence. Blood. 2017;129(12):1595-606.
2. Huang Y, Sanguinetti G. BRIE: transcriptome-wide splicing quantification in single cells. Genome biology. 2017;18(1):123.
3. Choi H, Sheng J, Gao D, Li F, Durrans A, Ryu S, et al. Transcriptome analysis of individual stromal cell populations identifies stroma-tumor crosstalk in mouse lung cancer model. Cell reports. 2015;10(7):1187-201.
4. Giustacchini A, Thongjuea S, Barkas N, Woll PS, Povinelli BJ, Booth CAG, et al. Single-cell transcriptomics uncovers distinct molecular signatures of stem cells in chronic myeloid leukemia. Nat Med. 2017;23(6):692-702.


Using genetics to investigate the increased risk of cardiovascular disease in people suffering from serious mental illness

Prof Daniel J. Smith, Institute of Health and Wellbeing, University of Glasgow
Dr Rona J. Strawbridge, Institute of Health and Wellbeing, University of Glasgow
Dr Hanna M. Björck, Department of Medicine Solna, Centre for Molecular Medicine, Karolinska Institutet, Sweden

PhD Project Summary:

Epidemiology has convincingly demonstrated that people suffering from serious mental illness, have an excess risk of cardiovascular diseases (CVD) 1. Serious mental illness frequently presents with low socioeconomic status, sedentary lifestyles, poor nutrition and smoking, all of which are recognised risk factors for CVD. It remains unclear whether the excess of CVD in people with SMI is a result of their increased CVD risk burden, or whether there are shared biological mechanisms for SMI and CVD.

Two major causes of cardiovascular mortality are atherosclerosis and venous thromboembolism (VTE). Atherosclerosis is characterised by vascular remodeling resulting in reduced elasticity of blood vessels, reduced lumen size, lipid deposition in the vessel wall and increased susceptibility to endothelial damage. Damage or rupture of the vascular wall causes thrombus formation, which can occlude the blood vessel causing myocardial infarction or stroke. Atherosclerosis is a progressive and systemic process, which proceeds cardiovascular events by decades. Non-invasive imaging of the intima-media thickness (cIMT, reflecting vascular remodeling) of the carotid artery is an accepted method for assessing sub-clinical atherosclerosis. Venous thromboembolism (VTE) accounts for a smaller proportion of CVD but is frequently fatal. VTE is an acute event resulting from genetic predisposition and an environmental trigger. Predisposition includes genetic variations which enhance the thrombotic potential (such as reduced protein S levels). Environmental triggers include conditions whereby blood flow to extremities is reduced, such as immobilization (including catatonia or sedation to minimize psychotic episodes). VTE encompasses deep vein thrombosis, where a thrombus occludes a vein in the lower limb, and pulmonary embobolism, where a fragment of a thrombus in the circulation causes occlusion of lung vasculature. Patients presenting with VTE typically have no indications of risk prior to the event, therefore it is hard to predict or prevent.

This project aims to:

  • Determine whether genetic predisposition to serious mental influences CVD, specifically atherosclerosis and venous thrombosis. Genetic loci associated with schizophrenia, bipolar disorder and major depressive disorder will be tested for influence on cIMT measures in the IMPROVE 2 and UK biobank cohorts 3 (total n=5,370). Replication in additional cohorts with cIMT measures and a subsequent meta-analysis will be conducted (expected n~60,000). The loci will also be assessed for impact on risk of VTE, in the UK biobank cohort (n cases=3,290 and n controls=116,868 (interim release)) with replication in additional VTE cohorts (anticipated n cases= 24,000 and n controls=520,000 (final release)). The potential to use genetic predisposition to serious mental illness as a tool for stratification of CVD risk will also be assessed.
  • Identify genes through which genetic variants are acting. Loci which influence either cIMT or VTE will be interrogated for effects on gene expression in relevant tissues (peripheral blood, arterial and endothelial tissues) using the GTEx (publically available, n=119-338) and ASAP/DAVACCA datasets (n=600-1600) 4. Cis and trans effects will be analysed in both datasets, with analysis of allelic imbalance also being conducted in ASAP/DAVAACA. 
  • Elucidation of mechanisms by which genetic predisposition to serious mental illness impacts CVD. The results of all analyses will be compiled and combined with bioinformatics analyses to map out the mechanisms for the increased atherosclerosis and risk of VTE, in the context of the current understanding of mental health and CVD.

Training outcomes:
Through this project the PhD student will gain a solid understanding of considerations and limitations in both statistical and biological aspects of complex disease research. Competence with robust statistical genetics, standard epidemiology, gene expression analysis and interpretation of data are the key areas where specific training will be provided. Critical reading of research papers, collaboration etiquette, writing and presentation skills are all fundamental to a research career so will be continually developed throughout the project.


  1. M DEH, Correll CU, Bobes J, Cetkovich-Bakmas M, Cohen D, Asai I et al. Physical illness in patients with severe mental disorders. I. Prevalence, impact of medications and disparities in health care. World Psychiatry 2011; 10(1): 52-77.
  2. Baldassarre D, Nyyssonen K, Rauramaa R, de Faire U, Hamsten A, Smit AJ et al. Cross-sectional analysis of baseline data to identify the major determinants of carotid intima-media thickness in a European population: the IMPROVE study. Eur Heart J 2010; 31(5): 614-622.
  3. biobank U. Genotyping of 500,000 UK Biobank participants. Description of sample processing workflow and preparation of DNA for genotyping. 2015; 11 September 2015.
  4. Folkersen L, van't Hooft F, Chernogubova E, Agardh HE, Hansson GK, Hedin U et al. Association of genetic risk variants with expression of proximal genes identifies novel susceptibility genes for cardiovascular disease. Circ Cardiovasc Genet 2010; 3(4): 365-373.

Using prescribing preferences as an instrumental variable to obtain causal estimates of treatment effects from non-randomised studies

Dr David A McAllister, Institute of Health & Wellbeing, University of Glasgow
Dr Jim Lewsey, Institute of Health & Wellbeing, University of Glasgow
Dr Nicholas Mills, Centre for Cardiovascular Disease, University of Edinburgh
Prof David Newby, Edinburgh Wellcome Trust Clinical Research Facility, University of Edinburgh

PhD Project Summary

In this project, the student will develop expertise in data science. They will analyse large complex datasets using up to date statistical techniques to answer clinically important questions. Specifically, they will explore the use large routine healthcare databases to address questions which cannot feasibly be addressed using randomised clinical trials (RCTs).

RCTs are the gold standard for obtaining causal estimates of treatment effects as they account for both measured and unmeasured confounding. However, it is not feasible to conduct sufficiently large clinical trials to determine treatment effects within precisely-defined sub-groups, even for the most pressing clinical questions. Therefore, evidence as to treatment effectiveness is often reduced to one-size fits all, i.e. the average treatment effect.

There has been considerable interest in analysing observational data to overcome this limitation. However, this has been undermined by the problem of confounding by indication; patients offered treatments differ from those not offered treatments, making comparisons unreliable. In attempts to overcome this problem, a number of methods have been explored. Of these, instrumental variable (IV) analysis is promising as, provided a good IV can be identified (i.e. a variable which is associated with the outcome only via its association with the treatment allocation), it can produce unbiased estimates of treatment effects.

For example, in a study of invasive vs non-invasive management of acute myocardial infarction, regional cardiac catheterisation rate was used as an IV. In the IV analysis, the relative risk reduction (RRR) for mortality was 16%, which is comparable to estimates from a number of RCTs (RRR 8 – 21%). In contrast, the results from regression-based methods employed to control for confounding by indication (multivariable regression and propensity scores) substantially overestimated the treatment benefit (RRR 50%) [1].

This research will explore the potential of “prescribing preference” as an IV, which has been shown to have good properties in the literature [2]. The effectiveness of ticagrelor and prasugrel compared to clopidogrel for patients with acute coronary syndrome will be used as an exemplar, as it has been recommended for most patients with the condition, with the proviso that certain sub-groups may not benefit (SIGN 148

Data suitable for this question will be obtained from ongoing RCTs embedded within routine healthcare data [4]. In addition to data obtained from the Scotland-wide databases (hospitalisation, death, prescribing etc.), this cohort includes angiography and electrocardiography data, as well as primary data on the acute event (heart rate, blood pressure, etc.) and adjudicated data on major cardiovascular events and death.

Prescribers will be identified using data from these cohorts, and their preference for ticagrelor/prasugrel compared to the older agent clopidogrel will be characterised according to patient characteristics and calendar time. The primary outcome will be recurrent myocardial infarction or cardiovascular death at 1 year. Different definitions of prescriber preference will be used to construct a range of IVs (e.g. defined by last 10 patients prescribed) which will be subsequently evaluated. The performance of each IV will then be compared to results from large, well conducted RCTs [3]. We will also use regression-based methods to estimate the treatment effects (regression modelling, propensity scores) and compare these to the IV approach. Having validated the use of prescriber preference as an IV for the entire cohort, we will explore the use of this approach in precision medicine by estimating treatment effects for patient sub-groups.

1) to test whether prescriber preference is a suitable IV for obtaining a causal estimate of ticagrelor/prasugrel compared to clopidogrel in this population; 2) to compare the properties and performance of differently constructed IVs, 3) to explore the use of IVs in estimating treatment effects for sub-groups.

Training outcomes:
Skills will be learnt which are crucial to modern data science practitioners, such as collaborating with others from a range of disciplines, the use of version control software and literate programming, data visualisation, and the streamlined production of dynamic documents and web apps


  1. TA Stukel, ES Fisher, DE Wennberg, et al. Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA 2007; 297: 278–28.
  2. NM Davies, D Gunnell, KH Thomas, et al. Physicians’ prescribing preferences were a potential instrument for patients’ actual prescriptions of antidepressants. Journal of Clinical Epidemiology 2013; 66: 1386-1396.
  3. L Wallentin, et al. “Ticagrelor versus Clopidogrel in Patients with Acute Coronary Syndromes.” NEJM 361; 11:1045–57. doi:10.1056
  4. NL Mills, et al. High-Sensitivity Troponin in the Evaluation of Patients With Acute Coronary Syndrome: A Randomized Controlled Trial. ID NCT01852123

University of Edinburgh Projects

University of Edinburgh Projects

The MRC DTP in Precision Medicine is funded jointly with the University of Edinburgh, in addition to the Glasgow projects listed above, Edinburgh also have a number of PhD projects which you can read more about here: