Tensor decomposition based feature extraction and classification to detect natural selection from genomic data.

Area of Science:

• Evolutionary biology
• Genomics
• Bioinformatics

Background:

• Detecting natural selection is crucial for understanding trait evolution and adaptation, like lactose digestion or viral variant spread.
• Traditional methods for identifying selection footprints from genomic data are limited by simple models and patterns.
• Machine learning, particularly convolutional neural networks, shows promise but faces challenges like parameter estimation and feature localization.

Approach:

• This study introduces T-REx (Tensor REgulation extraction), a novel framework leveraging tensor decomposition to extract features from haplotype data.
• T-REx preserves the latent structure of multidimensional genomic data before feeding features into classical machine learning models.
• The approach focuses on extracting meaningful features from haplotype images for subsequent analysis.

Key Points:

• T-REx demonstrates high power and accuracy in discriminating between selective sweeps and neutral evolution in simulated scenarios.
• The method is robust to common technical challenges encountered in genomic data analysis.
• T-REx facilitates easy visualization of feature importance, aiding in the interpretation of results.

Conclusions:

• T-REx offers a powerful and versatile addition to the computational toolkit for detecting adaptive processes in genomic data.
• The integration of tensor decomposition with machine learning provides a robust approach for analyzing complex evolutionary patterns.
• This framework enhances our ability to study the genetic basis of adaptation across diverse biological systems.