A method to build extended sequence context models of point mutations and indels.

Area of Science:

• Genetics and Genomics
• Computational Biology
• Bioinformatics

Background:

• Mutation rates in the human genome are influenced by surrounding sequence context (k-mers).
• Traditional k-mer modeling is limited by data sparsity for larger k-mer sizes.

Purpose of the Study:

• To introduce a novel method, k-mer pattern partition (kPaP), for more accurate mutation rate modeling.
• To develop software (kmerPaPa and Genovo) for implementing kPaP and predicting mutation types.

Main Methods:

• Grouping similar k-mers to overcome data sparsity issues in mutation rate modeling.
• Utilizing a large dataset of human de novo mutations for model training and validation.
• Developing the kmerPaPa software for k-mer pattern partition and the Genovo software for functional mutation prediction.

Main Results:

• kPaP significantly improves the prediction of mutation rates, enabling models with wider sequence contexts.
• The method accurately predicts rates for point mutations, insertions, and deletions.
• Genovo, using kPaP models, enhances statistical power for detecting disease-causing variants and identifying genes under selective constraint.

Conclusions:

• K-mer pattern partition offers a robust solution for modeling sequence context-dependent mutation rates.
• The developed software facilitates advanced analysis of mutation patterns and their functional implications.
• This approach advances our ability to identify disease-associated genes and understand evolutionary constraints.