A foundation model for learning genetic associations from brain imaging phenotypes

Due to the intricate etiology of neurological disorders, finding interpretable associations between multiomics features can be challenging using standard approaches.

We propose COMICAL, a contrastive learning approach using multiomics data to generate associations between genetic markers and brain imaging-derived phenotypes. COMICAL jointly learns omics representations utilizing transformer-based encoders with custom tokenizers. Our modality-agnostic approach uniquely identifies many-to-many associations via self-supervised learning schemes and cross-modal attention encoders. COMICAL discovered several significant associations between genetic markers and imaging-derived phenotypes for a variety of neurological disorders in the UK Biobank, as well as prediction of diseases and unseen clinical outcomes from learned representations.

The source code of COMICAL along with pretrained weights, enabling transfer learning, is available at https://github.com/IBM/comical.