Project Reference: TRG27-SAN-JN1

Project title: A generative somatic genotype-methylome-to-phenotype model of clonal evolution in blood

Project outline

Blood stem cells accumulate somatic mutations and methylation changes over life. Clonal haematopoiesis is near-universal in older people and associated with blood cancers, cardiovascular disease, and other ageing phenotypes. Here we wish to explore if we can predict a cell's fitness, expansion trajectory, and disease outcome from its somatic and epigenetic changes. This would make somatic evolution predictive, enabling risk stratification and earlier intervention.

Existing variant-effect models target germline variation — protein/expression effects, not competitive fitness within a tissue over decades. None trains on the somatic genome at single-cell resolution, since no dataset previously linked acquired (epi)mutations to downstream clonal fitness in vivo. This project uses such a dataset.

Our in-house cohort of >1000 haematopoietic colonies spans donors across human lifespan, healthy and diseased, from indolent clonal haematopoiesis to myeloproliferative neoplasms and explosive BCR::ABL1-driven CML. Each colony has five linked layers: dated somatic phylogenies from whole-genome sequence; whole-genome methylome at single-CpG resolution; phylogeny-derived fitness from near-neutral drift to >30,000%/year CML growth; developmental timing of each change; and clinical annotation.

This is the only dataset jointly measuring genotype, methylome, and phylogeny-derived fitness at clonal resolution across health and cancer. You will build a genomic foundation model with a methylation-aware input and fitness head, supervised by clonal growth rates rather than curated labels, and then apply it to in silico driver discovery, benchmarked against linear baselines.

The project spans clinical/computational haematology, methylation genomics, and ML — training the student in foundation-model development, multimodal data curation, and benchmarking, extending prognosis tools toward a general predictor of somatic fitness for any tissue with measurable clonal selection.

Principal Supervisor

Dr Jyoti Nangalia

Email address: jn5@sanger.ac.uk

Department / Partner Institute: Sanger Institute

Website: https://www.sanger.ac.uk/person/nangalia-jyoti/ or https://www.nangalia.org/

BBSRC strategic theme

Transformative technologies