AI-driven stratification of missense variation in different disease contexts

Human disease is complex and multifaceted, with many underlying causes linked to genetic and somatic mutations, pathogens and interactions with our environment.

Most computer-based approaches for the prediction of disease-associated genetic variation assume the missense mutations involved are directly pathogenic in some way.

The reality is disease is highly complex, involving diverse molecular mechanisms and the interplay of many genes, such that illness is very much an emergent property linked to our dynamic molecular systems.

Addressing this gap will require new ways to characterise missense mutations at a system level including genetic background, interactions with other key molecules, and the broader evolutionary and pathogenic context of the associated disease.

This project will develop a novel computational framework, leveraging state-of-art artificial intelligence (AI) tools such as large language models trained for proteins, advanced structural prediction methods, and cutting-edge data, to categorize and understand the molecular variations associated with different disease types, including hereditary conditions, cancer driver mutations and infectious diseases.

We aim to stratify variation into functional consequences by using AI to analyze protein structures and their interactions.