Identification of natural selection in genomic data with deep convolutional neural network

Area of Science:

• Genomics
• Computational Biology
• Evolutionary Biology

Background:

• Genomic datasets are growing, making information extraction complex.
• Supervised Machine Learning (SML) and Convolutional Neural Networks (CNNs) show promise for analyzing demographic and adaptive processes in genomic data.
• The potential of SML in evolutionary genomics requires further exploration.

Purpose of the Study:

• To propose and evaluate an SML method for classifying genomic data windows.
• To utilize CNNs for detecting signatures of natural selection within genomic sequences.
• To assess the model's predictive accuracy on simulated and real population genomic data.

Main Methods:

• Genomic data represented as sequence windows from a population sample.
• Application of a Convolutional Neural Network (CNN) for classification.
• Training the model on simulated data to predict neutral and selection processes.

Main Results:

• The proposed SML method accurately classifies genomic windows.
• The CNN model effectively identifies signatures of natural selection.
• The model achieved nearly 90% accuracy in predicting neutral and selection processes.

Conclusions:

• SML, particularly CNNs, offers a powerful approach for evolutionary genomics.
• The developed method accurately detects natural selection in genomic data.
• This approach can handle large genomic datasets for population variability analysis.