Interpreting Convolutional Neural Networks in Population Genetics.

CNNs in population genetics implicitly learn summary statistics like pairwise heterozygosity and efficiently approximate long-range linkage disequilibrium, outperforming traditional methods in some aspects.

• Why it matters: Understanding what CNNs learn is crucial for validating their use in population genetics and for gaining biological insights, addressing the interpretability gap in deep learning applications.
• What they did: The study analyzed CNNs trained for data discrimination and detecting selective sweeps, using correlation analysis, SHAP values, dimensionality reduction, and interpretable models like decision trees and random forests.
• The result: Findings show CNNs can implicitly compute key summary statistics and efficiently model complex features such as linkage disequilibrium, enhancing interpretability and potential for biological inference.