Genetic Medicine PhD: How Do Multiple Missense Variants in One Protein Affect Its Function?

Genetic Medicine PhD: How Do Multiple Missense Variants in One Protein Affect Its Function?

Quick Decision

Clinical variant interpretation usually scores missense variants one at a time. This MRC FGx-DLA project will develop a predictor for the combined effect of multiple missense variants in the same protein or complex, then test it against structural, population-genetic, clinical and functional-assay data. Planned validation includes double-mutant MAVE data, known pathogenic variants in cis and phased cohort sequences. The proposed cross-site team spans Exeter rare-disease interpretation, Edinburgh benchmarking and Imperial protein-structure modelling.

The project suits applicants interested in computational biology, protein structure, machine learning or clinical genomics. Public materials name several large cohorts but do not specify the student’s access, sample counts, phasing resolution, benchmark design or route to clinical use. Exeter’s current MRC pages list 5 January 2027; a captured job record gives 8 December 2026 and mislabels the funder as EPSRC. Confirm the governing date and international-fee terms in writing; UKRI does not cover visa or IHS costs.

Research Problem and Doctoral Work

What is unsolved and why it matters

Missense variants replace one amino acid with another. The project description says variants of uncertain significance (VUS) are growing rapidly in clinical databases and that more than 70% of VUS are missense. It also notes that an average human genome contains around 10,000 missense variants and that about 1,000 genes per person have more than one. Despite this, clinical interpretation still commonly considers variants individually. Existing computational approaches can model multiple substitutions, but the project frames many such methods as designed for protein engineering rather than clinical interpretation. These figures and the problem framing come from the official project PDF.

The core scientific question is whether the effect of a variant can be predicted more accurately when the presence and position of another variant are represented explicitly. A pair can behave differently from the sum of two independently estimated effects. That makes variant phase, protein structure, interaction context, isoform choice and the biological assay used for validation central to the work. “Effect prediction” also needs to be distinguished from a clinical pathogenicity verdict: the project aims to help interpretation, not to replace clinical assessment or provide diagnosis by itself.

Advertised tasks, methods and expected output

The official project description outlines a staged computational project:

1. Review and benchmark existing multi-substitution approaches, including evolutionary methods, protein language models and three-dimensional structure-based predictors.
2. Develop predictions for one missense variant in the presence of another across four settings: within one structured domain; across domains in the same protein; in disordered or inter-domain regions; and across different proteins in the same complex.
3. Incorporate features such as three-dimensional proximity, amino-acid interactions, substrate binding, post-translational modifications, domain stability, conservation and constraint.
4. Test whether predictions capture interaction effects beyond the individual variants using existing double-mutant multiplexed assays of variant effect (MAVEs), known pathogenic alleles in cis, and population-cohort sequencing with statistical or read-based haplotyping.

The possible extension to three or more variants is explicitly optional. The intended main outcome is a new “multiVEP” for multiple missense variants in the same molecule, validated against experimental and clinical data and made available to the research community. This is an intended project outcome, not a guarantee of clinical deployment or regulatory approval.

The project PDF names UK Biobank, All of Us, Our Future Health and Genomics England as available data resources, and notes existing high-performance computing and cloud support. Those are useful starting points for population-scale analysis, but public project information does not specify the access approvals, project-level sample sizes, permitted linkage, compute allocation or data release conditions. Confirm these before treating a particular cohort or pipeline as secured.

Evidence-backed research angles and technical limits

Current single-variant systems show why the project is timely, while also setting a high bar for validation. ESM1b-based prediction has produced broad single-substitution scores across human protein isoforms; AlphaMissense combines structural context and evolutionary information for proteome-wide missense prediction; and Missense3D assesses structural changes on individual proteins and protein complexes. These tools provide strong comparison baselines, but their single-variant scores do not by themselves answer whether two phased changes interact in one molecule. The Atlas of Variant Effects predictor guide also emphasizes matching a predictor to the question it was built to answer.

A promising analytical design would separate additive and interaction effects: first establish a reproducible baseline from each variant scored alone, then test whether pair-specific terms improve prediction across the four advertised structural contexts. Evaluation should prevent leakage between training and test variants, genes, assays or homologous proteins. It should report performance by context and mechanism rather than one aggregate score. The 2025 VEP release recommendations specifically highlight transparent training/test data, leakage control and clear output limits. These are analysis suggestions based on the literature, not additional advertised requirements.

MAVE data are valuable because they can measure many variants experimentally, but they are assay-specific. The 2024 MAVE reporting recommendations call for adequate description of constructs, model systems, measurement and data processing. The 2025 Livesey–Marsh analysis finds that interpretation depends on using diverse assay data; agreement with a functional assay does not make that assay clinically representative for every gene or disease. Population-cohort validation introduces a different challenge: statistical phase and read-based haplotyping have different resolution, and population associations may not provide direct biochemical evidence. A strong project will state which endpoint each dataset supports, and where a prediction is about molecular function rather than disease risk.

School, Programme and Field Context

The project is officially listed through the University of Exeter and its Department of Clinical and Biomedical Sciences. Exeter’s research-degree page lists an MPhil/PhD in Clinical and Biomedical Sciences, full-time and part-time modes, and campuses including Streatham and St Luke’s. That is a plausible registration route, but the project document does not state the specific registration route, campus or degree-awarding arrangement for this cross-institutional project. Confirm those details with PGR admissions. Edinburgh and Imperial are identified as supervisory partners; this does not establish a joint or double degree.

Exeter’s Clinical and Biomedical Sciences department has relevant strengths in genomics, rare monogenic conditions, data science, AI, statistics and high-performance computing. Its postgraduate research page describes close work with NHS and NIHR research partners and lists access to facilities in genetics, long-read sequencing, bio-imaging and flow cytometry. These are department-level research strengths and do not mean that every facility will be used by this primarily computational project.

The school reports being Top 10 in the UK for world-leading and internationally excellent Clinical Medicine research, based on 4-star plus 3-star REF 2021 research; it also lists top-125 Life Sciences and top-175 Medical and Health subject positions in the Times Higher Education World University Rankings 2026, plus a top-200 combined Life Sciences and Medicine position in the same subject rankings. Separately, Exeter’s REF 2021 results page reports 45% 4-star and 48% 3-star in Clinical Medicine (UoA 1), and 41% 4-star and 49% 3-star in Biological Sciences (UoA 5). These are subject and institutional research signals, not a ranking or quality rating of this specific PhD.

The MRC FGx-DLA is a national four-year programme across Exeter, Cambridge, Edinburgh, Imperial, King’s and Oxford. Its postgraduate pages describe inter-institutional supervision, shared cohort training, scientific meetings, networking and cross-site collaboration. The national programme lists training in functional genomics, statistics, bioinformatics, AI and machine learning, research integrity, leadership, public engagement and career development. The exact training calendar and the division between central and project-specific supervision should be requested before accepting an offer.

Supervisor and Lab Assessment

Supervisor research directions

The advertised division is coherent: Caroline Wright and Joseph Leslie at Exeter bring rare-disease clinical interpretation and gene–disease discovery; Joe Marsh at Edinburgh contributes VEP comparison and validation; Alessia David at Imperial contributes protein-structure-based predictor design. This is a credible combination for a project that must bridge clinical genomics, benchmarking and structural bioinformatics. Wright’s official profile describes work on rare-disease diagnosis, penetrance, variant filtering and in-silico structural modelling. Exeter’s Translational Genomics group works with large-scale sequencing and multi-omics to improve rare-disease diagnosis, including UK Biobank, Genomics England and DECIPHER. David’s profile describes co-development of Missense3D, Missense3D-DB and PhyreRisk. Marsh’s laboratory focuses on VEP evaluation, disease mechanisms and multiplexed assays.

Recent important publications

• Livesey and Marsh (2025), “Variant effect predictor correlation with functional assays is reflective of clinical classification performance.” The study analyzes how VEP and functional-assay performance relate to clinical classification, and why a range of assays is more informative than reliance on one or a few. It directly informs the project’s benchmark design, but it is not a demonstration that a multi-variant predictor has already been solved.
• Badonyi and Marsh (2025), “Prevalence of loss-of-function, gain-of-function and dominant-negative mechanisms across genetic disease phenotypes.” This work makes mechanism distinctions central to variant interpretation. For multiVEP, a general deleteriousness score may be insufficient where different molecular mechanisms produce different disease outcomes.
• Kingdom et al. (2024), “Genetic modifiers of rare variants in monogenic developmental disorder loci.” Using population cohort data, Wright and colleagues report that additional rare and common variants can modify phenotypic effects. This supports contextual, background-aware interpretation, but it does not establish pairwise epistasis between two missense variants in one protein.
• Tsitsa et al. (2026), “Missense3D: leveraging protein structures, including AlphaFold models, to understand the effect of missense variants.” David and co-authors describe updated analysis of protein structures and protein–protein complexes, including automated sequence-to-structure mapping and quality assessment. This is relevant structural expertise, not a promised project tool or deliverable.

These papers make the team’s skill mix credible. Applicants should ask how the supervisors will coordinate cross-site work, who will lead day-to-day guidance, and how much time is planned for clinical data, structural modelling and benchmarking.

Funding, Application, and Living Cost

Award and fee terms

The current Exeter MRC FGx-DLA programme and Ref. 6025 award page list September 2027 entry, a four-year studentship, UK tuition fees, a UKRI minimum stipend of £21,805 per year for 2026/27 (updated annually), and up to £5,000 per year for research training and support. Part-time study is available. The project-level source correction confirms that the captured “EPSRC” label was wrong: this is an MRC FGx-DLA project.

The award page describes coverage as “UK tuition fees,” while the UK Human Functional Genomics Initiative postgraduate page says that Exeter does not require international students to supplement UKRI funding with the international fee difference. UKRI’s general guidance says it contributes at the home-fee rate and that overseas students may need to pay a difference, so the programme-specific Exeter statement is important. Ask the admissions team to confirm that the no-top-up rule applies to this project and that it will appear in the offer letter. UKRI says visa fees and the Immigration Health Surcharge (IHS) are not covered.

The University’s April 2026 estimate for basic costs in Exeter is £1,300–£2,000 per month, depending on housing and lifestyle. Dividing the 2026/27 stipend by 12 gives about £1,817 per month. This is a rough comparison, not a personal budget: at the lower end there may be room for other costs, while at the higher end the stipend alone would not cover the estimate. Initial housing deposits, travel, visa/IHS, equipment and exchange-rate changes can add pressure. Request the actual stipend schedule and estimate housing costs before deciding.

Deadline, eligibility and application materials

The official MRC FGx-DLA programme and award pages give 5 January 2027 as the deadline for September 2027 entry. The captured job listing instead gives 8 December 2026 and incorrectly identifies EPSRC. As the listing may reflect an earlier date or a stale error, treat 8 December as a conservative preparation target and ask PGR admissions to confirm the deadline and time zone. Do not present 8 December as verified for this exact award.

Programme eligibility is a minimum 2:1 honours degree (or international equivalent) in a relevant subject. Listed fields include genetics, molecular and biomedical sciences, human biology, biochemistry, medicine, healthcare professions, bioinformatics, computer science and statistics. The typical English-language minimum is IELTS 7.0 overall with no component below 6.5, or equivalent. The application has two required parts: the Exeter online application and FGx-DLA Applicant Questionnaire. It asks for transcripts, a maximum two-page CV, two references and English-language evidence where required. Applicants can list up to three project preferences but should submit only one application; duplicate applications are treated as unsuccessful after the first. Referees should not be members of the prospective supervisory team. If shortlisted, referees must send references seven days before the interview; the applicant must arrange this.

The official criteria do not list a particular programming language, protein-structure package or machine-learning background as a separate admission prerequisite. Those are useful preparation areas, not extra formal requirements.

Applicant Fit and Preparation

The project is best suited to applicants with a solid base in genetics, computational biology, bioinformatics, statistics, computer science, biomedical science or a related life-science discipline, and interest in building a model that is evaluated against experimental and clinical data. It may be less suitable for someone primarily seeking wet-lab experimentation, a narrow protein-structure problem without cohort or clinical interpretation, or a project with a fixed model architecture and dataset already specified.

Useful preparation to evidence in the application could include Python or R data analysis, reproducible benchmarking, machine learning, sequence/protein language models, protein structure and variant annotation, statistical genetics or haplotype analysis. Applicants from biology/medicine can emphasize quantitative coursework and data projects; applicants from computing can show genomics or protein biology exposure. The panel’s formal baseline is the relevant degree and English-language criteria, not every skill mentioned in this report.

A concise application should explain why multi-variant context matters for rare-disease interpretation, how previous work demonstrates the ability to learn across disciplinary boundaries, and how the applicant approaches validation and reproducibility. The DLA questionnaire explicitly asks about motivation, a meaningful achievement, lessons from prior research experience, and transparent use of generative AI if applicable. A useful interview preparation exercise is to compare the assumptions of a sequence model, a structural predictor and a clinical-label predictor, then identify how to test for data leakage and assay mismatch.

Next Actions and Questions to Verify

Before applying

1. Ask Exeter PGR admissions to confirm the 5 January 2027 date, submission cut-off time and time zone, and whether the earlier 8 December 2026 date has any effect on this exact project.
2. Confirm that the application is for MRC FGx-DLA Ref. 6025, and that the national initiative’s no-international-fee-top-up rule applies to this project. Ask how the no-gap rule will appear in the offer and what international applicants must budget for visa and IHS.
3. Confirm that the degree and research registration route will be Exeter MPhil/PhD Clinical and Biomedical Sciences, the campus/base location, and whether any formal registration or activity is required at the Edinburgh or Imperial partner sites.
4. Prepare one application with the Exeter online form and questionnaire, transcripts, two-page CV, two external references and English evidence as applicable. Check the rule allowing up to three project preferences.

At interview

• Which datasets are already approved for this project, and what are the expected sample sizes and data-use restrictions for each of UK Biobank, All of Us, Our Future Health and Genomics England?
• Which double-mutant MAVEs and known cis pathogenic examples are currently usable, and how will assay context and disease mechanism be assessed?
• What are the planned benchmark splits to prevent overlap by variant, gene, protein homology or assay between development and testing?
• How will the team define clinically useful validation without presenting a molecular effect score as a standalone pathogenicity classification?
• What is the likely progression between pairwise modeling and the optional 3+ variant extension?
• Who will provide day-to-day supervision, how often will the three institutions meet, and what cohort training is compulsory?

Before accepting an offer

• Obtain written confirmation of the fee coverage, stipend amount and annual update, £5,000/year research-support arrangements, start/end dates, registration route, campus and international-student costs.
• Ask for the agreed access/compute plan and how delays in cohort permissions, phasing or assay data would affect the project milestones.

Evidence and Sources

Core project, programme and school

• Official project PDF
• MRC FGx-DLA programme and application page
• Exeter award Ref. 6025
• UK Human Functional Genomics postgraduate programme and Exeter international fee guidance
• Exeter Clinical and Biomedical Sciences research degrees
• Clinical and Biomedical Sciences research themes
• Exeter REF 2021 results

Supervisors and recent publications

• Exeter Translational Genomics
• Caroline Wright, NIHR Exeter BRC
• Exeter Rare Disease Genomics team
• Joe Marsh, University of Edinburgh
• Alessia David, Imperial College London
• Livesey and Marsh (2025), VEP and functional-assay benchmark
• Badonyi and Marsh (2025), disease mechanisms
• Kingdom et al. (2024), rare-variant genetic modifiers
• Tsitsa et al. (2026), Missense3D update

Field, methods and clinical evidence

• Missense3D structural analysis tool
• Atlas of Variant Effects predictor guide
• Guidelines for releasing a variant effect predictor (2025)
• ESM1b proteome-wide prediction study (2023)
• AlphaMissense study (2023)
• MAVE reporting guidelines (2024)
• ProteinGym benchmark (NeurIPS 2023)
• UK practice guidelines for variant classification (ACGS, 2024)

Funding, cost and careers

• UKRI-funded student support, stipend and visa/IHS scope
• Exeter student budgeting estimate
• Skills England 2026 Life Sciences Sector Skills Needs Assessment
• UK Life Sciences Jobs Plan (2026)