An automated multi-modal graph-based pipeline for mouse genetic discovery.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genome-Wide Association Studies (GWAS) often yield numerous false positives, hindering the identification of true causative genetic factors in mouse models.
• Accurate identification of genetic factors is crucial for understanding human diseases and developing effective treatments.

Purpose of the Study:

• To develop an automated pipeline (GNNHap) utilizing graph neural networks (GNNs) to enhance the accuracy and speed of identifying causative genetic factors in mouse genetic models.
• To improve upon existing methods for genetic discovery by integrating diverse data sources.

Main Methods:

• Developed GNNHap, a graph neural network-based automated pipeline for analyzing mouse genetic model data.
• Assessed allelic associations with strain response patterns and analyzed 29 million published papers for gene-phenotype relationships.
• Incorporated protein-protein interaction networks and protein sequence features into the GNN analysis.

Main Results:

• GNNHap significantly improved results compared to simple linear neural networks.
• Identified novel causative genetic factors for murine models of diabetes/obesity and cataract formation.
• Validated findings through gene knockout mouse phenotypes, demonstrating the pipeline's efficacy.

Conclusions:

• GNNHap effectively identifies high-probability causative genetic factors, accelerating genetic discovery in complex traits.
• The characterization of genetic architecture in murine models facilitates the development of precision medicine approaches for new therapies.