MuRaL-indel: a deep learning framework for building insertion and deletion mutation rate maps.

Area of Science:

• Genomics
• Computational Biology
• Population Genetics

Background:

• Germline insertions and deletions (INDELs) are significant genetic variants impacting genome evolution and disease.
• Accurate quantification of fine-scale INDEL mutation rates is hindered by data limitations and variant diversity.

Purpose of the Study:

• To introduce MuRaL-indel, a deep learning framework for predicting germline INDEL mutation rates.
• To generate high-resolution, base-specific mutation rate maps for the human genome and other species.
• To improve the identification of pathogenic INDELs and understand INDEL mutational mechanisms.

Main Methods:

• Developed MuRaL-indel, a deep learning framework utilizing a U-Net architecture for long-range sequence context.
• Applied the framework to extensive rare variant data from large population cohorts.
• Validated MuRaL-indel's performance against existing models and across multiple species.

Main Results:

• MuRaL-indel achieves superior accuracy in predicting INDEL mutation rates compared to existing models.
• Generated base-resolution, length-specific mutation rate maps for the human genome.
• Demonstrated broad applicability across human and non-human species (e.g., primates, insects, plants).
• Identified sequence motifs associated with increased INDEL mutability, offering mechanistic insights.
• MuRaL-indel-derived constraint scores improved prioritization of pathogenic INDELs.

Conclusions:

• MuRaL-indel provides a generalizable and scalable framework for high-resolution INDEL mutation rate mapping.
• This tool serves as a valuable resource for genome evolution, mutational mechanism studies, and variant interpretation in genetic disease research.