A mutation-level covariate model for mutational signatures

Area of Science:

• Genomics
• Computational Biology
• Population Genetics

Background:

• Understanding genome evolution requires analyzing mutational processes and their activity across the genome.
• Current methods often assume uniform mutational activity, neglecting potential influences of covariates like genomic region or DNA strand.

Purpose of the Study:

• To develop and validate the first mutation-covariate models that explicitly incorporate the impact of covariates on mutational process exposures.
• To assess the influence of replication strand on mutational processes and compare model performance against strand-oblivious approaches.

Main Methods:

• Development of novel mutation-covariate models incorporating specific covariates.
• Application of these models to analyze mutation data, focusing on replication strand effects.
• Comparative analysis of covariate-aware models versus standard, strand-oblivious models across diverse datasets.

Main Results:

• The proposed models successfully capture replication strand specificity in mutational processes.
• Identified specific mutational signatures influenced by replication strand.
• Models incorporating mutation-level covariate information demonstrated superior performance on held-out data.

Conclusions:

• Mutation-covariate models provide a more accurate framework for studying genome shaping mutational processes.
• Accounting for covariates like replication strand is crucial for a comprehensive understanding of mutation patterns.
• These advanced models enhance the predictive power and accuracy in genomic mutation analysis.