Undergraduates & Masters

Probabilistic machine learning and genomics - Dias & Frazer Lab

PROJECT DESCRIPTION - Folding versus function: disentangling the molecular mechanisms of missense variants in gene-trait associations

Protein language models | Missense variants | Protein stability | Protein folding | Human genetics | UK Biobank | Variant effect prediction | Computational biology | Machine learning | Protein biophysics

Genome sequencing has grown exponentially, but determining the phenotypic consequences of missense variants remains a key challenge in human genetics. Deep-learning unsupervised methods, such as protein language models (pLM), that learn the distribution of sequence variation across organisms have emerged as promising tools for scoring variant effects. In gene-trait association studies, regression tests using pLMs find ∼50% more associations than burden tests (Jang et al., Cell Genomics 2026) and can be run directly from precomputed summary statistics (Dinh et al., Nature Methods 2026). However, pLM are trained on evolutionary data and therefore often fall short of attributing effects to specific biophysical mechanisms.We propose to integrate protein language models with folding energetics to disentangle effects on stability and on functions beyond folding, such as allostery, ligand/partner binding and downstream interactions. We will quantify and separate mutational effects on the energetics of protein folding versus natural selection by computing protein “dark energy”, defined as the difference between physical folding free energies and evolutionary free energies derived from a pLM (Galpern et al., PNAS 2026).  We will apply this biophysical decomposition to study the genotype-phenotype summary statistics from the UK Biobank as provided by Genebass (Karczewski et al., Cell Genomics 2022), focusing specifically on genes associated with blood biochemistry biomarkers. We will relate the predicted folding-stability and dark energy changes for each missense variant to the effect sizes and contrast those regressions against the gene's putative loss-of-function (pLoF) effect. We will disentangle variants destabilizing the encoded protein from others that preserve protein stability but affect specific chemical activities. Together, this work will pave the way to a scalable framework to move beyond statistical gene–trait associations toward a residue-level, mechanistic understanding of how missense variation perturbs protein biology in human genetics.

WHO ARE WE LOOKING FOR?

We are looking for a motivated master student with a background in bioinformatics, computational biology, computer science, physics, mathematics, or a related quantitative discipline. Previous programming experience in Python is expected. Familiarity with molecular biology, genetics, statistics, or machine learning is desirable but not required. The student should be curious and interested in applying computational methods to biological questions.

HOW TO APPLY

Send applications documents to the PI and Ezequiel Galpern - HERE

Systems & Synthetic Biology - MARTIN Lab

PROJECT DESCRIPTION - Understanding the effect of environmental changes on a genotype-phenotype map

Evolution | Genotype-phenotype maps | Fitness seascapes

Models of evolution using genotype-phenotype maps and the resulting fitness landscapes allow for the accurate treatment of genetic sequences and the high-dimensional structure of mutations. Improvements in both computational resources and experimental methods make it possible to study increasingly large landscapes, but there are still many unanswered questions about what structures can and do exist in real genotype spaces, and how an individual's changing environment influences both its phenotype and fitness. This project will use computational tools to understand the effect of a changing environment on a genotype-phenotype map, and on the subsequent effect that a changing genotype-phenotype map has on evolutionary simulations.

WHO ARE WE LOOKING FOR?

As an interdisciplinary field, this project will be most interesting to a student either:
- With a physics/computer science background and an interest in learning more about biology and genetics, or 
- With a biology/bioinformatics background with an interest in learning more about programming and statistics. 
A strong motivation to learn about new topics is essential.

HOW TO APPLY

Send applications documents to the PI and Kye Hunter - HERE

Evolutionary Processes Modeling - Weghorn Lab

PROJECT DESCRIPTION - How do genetic and evolutionary factors shape the immunogenic potential of cancer-associated mutations

Evolution | Neoantigen | Immunogenicity

Cancer develops through the accumulation of somatic mutations, some of which generate neoantigens that can be presented by human leukocyte antigen (HLA) molecules and recognized by T cells. These interactions may influence which mutations persist during tumour evolution, but the extent and determinants of immune-mediated selection remain unclear. Progress has been limited by the difficulty of modelling mutation probabilities, antigen presentation, and T-cell recognition with sufficient accuracy. This project will investigate how genetic and evolutionary factors shape the immunogenic potential of cancer-associated mutations. Using comparative sequence analysis, evolutionary simulations, and context-aware computational models, we will examine how variation in coding sequences affects the likelihood that somatic mutations generate peptides with the potential for immune recognition. The work will provide a broader understanding of the relationship between genome evolution, somatic mutation, and tumour immunogenicity, and may help clarify the factors that influence immune selection during cancer development.

WHO ARE WE LOOKING FOR?

A student with: 
- An interest in computational biology, evolution, or cancer genomics.  
- Experience in Python (or similar) and handling large datasets.  
- Some familiarity with bioinformatics pipelines (preferred but not required).

HOW TO APPLY

Send applications documents to the PI, María Kelly and Francisco Javier Ordoñez - HERE

Contact

For any further questions, please contact

CRG Training & Academic Office
Centre de Regulació Genòmica
Dr. Aiguader, 88
PRBB Building
08003 Barcelona
training@crg.eu
 

Past opportunities

• CRG Summer Internship Programme
• CRG INTERNATIONAL MASTER INTERNSHIPS PROGRAMME