deepBreaks: a Machine Learning Tool for Identifying and Prioritizing Genotype-phenotype Associations

Abstract Sequence data (e.g., nucleotides or amino-acids) are critical for advancing our understanding of biology. However, there are many challenges with investigating and analyzing sequencing data and genotype-phenotype associations, including non-independent observations, large noise components, nonlinearity, colinearity, and high dimensionality. Therefore, machine learning (ML) algorithms are well suited for analyzing sequence data as they can capture nonstructural patterns of relationships with the biology of interest as genotype-phenotype associations. Nevertheless, flexible and user-friendly implementations of ML approaches are lacking, especially ones that take advantage of the unique features of high-volume DNA sequence data. Here, we present deepBreaks, a generic approach with unified data analysis steps that identify the most important positions in sequence data that correlate with phenotypic traits of interest. deepBreaks compares the performance of multiple ML algorithms and prioritizes the most important positions based on the best-fit models. deepBreaks is open-source software with documentation available online at https://github.com/omicsEye/deepBreaks.