Germline mutation rates are essential for genetic and evolutionary analyses. Yet, estimating accurate fine-scale mutation rates across the genome is a great challenge, due to relatively few observed mutations and intricate relationships between predictors and mutation rates. Here we present MuRaL (Mutation Rate Learner), a deep learning-based framework to predict fine-scale mutation rates using only genomic sequences as input. Harnessing human germline variants for comprehensive assessment, we show that MuRaL achieves better predictive performance than current state-of-the-art methods. Moreover, MuRaL can build models with relatively few training mutations and a moderate number of sequenced individuals. It can leverage transfer learning to build models with further less training data and time. We apply MuRaL to produce genome-wide mutation rate profiles for four species - Homo sapiens, Macaca mulatta, Arabidopsis thaliana and Drosophila melanogaster, demonstrating its high applicability. The generated mutation rate profiles and open-source software can greatly facilitate related research.
 

mutation rate,deep learning,genomics,bioinformatics,computational biology