The identification of damaging DNA variants is fundamental to the diagnosis, treatment and prevention of genetic disease. Computational methods for predicting variant effects are widely used for scoring and prioritising variants most likely to cause disease, but their current performance is not considered good enough to use as strong evidence in genetic diagnosis. Conversely, directly measuring the effects of variants using experiments can be powerful, but is time consuming and can be difficult to perform on a large scale. This project seeks to improve our ability to identify disease-causing variants through a combination of computational and experimental approaches, with a particular focus on the diverse molecular mechanisms by which mutations can cause disease, in contrast to the current overwhelming focus on simple loss-of-function effects.
In Aim 1, we will use structural bioinformatics techniques to explore the properties of dominant-negative, gain-of-function and loss-of-function variants, and learn how they are related to protein structure and disease phenotypes. In Aim 2, we will perform deep mutational scanning experiments (DMS) on human disease genes, enabling us to measure effects and elucidate molecular mechanisms for tens of thousands of single amino-acid substitutions. Finally, in Aim 3, we will seek to improve the prediction and prioritisation of pathogenic and other phenotypically important variants. This will involve assessing current approaches for predicting variant effects, implementing our own computational variant prioritisation pipeline and meta-predictor, and using our new understanding of molecular mechanisms to integrate computational predictors and DMS data with structural and other protein-level features. We will also demonstrate the power of our approach in application to sequencing data from clinical and population studies.
