AI + Genomics

High-throughput sequencing technologies led to the discovery of thousands of risk variants involved in various traits and diseases. However, conventional approaches are often incapable of unveiling intricate effects or relationships within complex and high-dimensional data. Leveraging explainable machine learning (ML) techniques to integrate multi-level data sources holds promise for establishing crucial functional links between associated variants and causal genes [1]. In the context of age-related macular degeneration [2], diverse genomic and gene expression features can be exploited to prioritize linked genes and illuminate relevant pathways. The ML approach serves as a solution for comprehending the causal genes involved. We thus propose the development of an ML framework aimed at learning the key characteristics of driver genes from genomic features, achieved through the integration of multi-level data and external validation.

[1] Novakovsky G, et al. Obtaining genetics insights from deep learning via explainable artificial intelligence. Nature Reviews Genetics. 2023 Feb;24(2):125-37.
[2] Fan Q, et al. Contribution of common and rare variants to Asian neovascular age-related macular degeneration subtypes. Nature Communications. 2023 Sep 11;14(1):5574.