Bridging the Generalization Gap: Training Robust Models on Confounded Biological Data

TL;DR

This paper presents methods to improve the generalization of biological data models by controlling confounders using normalization and adversarial training, demonstrated on simulated and real patient data.

Contribution

It introduces ONION normalization and DANN adversarial training to effectively reduce confounding effects in biological data modeling.

Findings

Significant improvement in model generalization on simulated data

Enhanced prediction accuracy on empirical patient data

Effective removal of confounder influence in biological datasets

Abstract

Statistical learning on biological data can be challenging due to confounding variables in sample collection and processing. Confounders can cause models to generalize poorly and result in inaccurate prediction performance metrics if models are not validated thoroughly. In this paper, we propose methods to control for confounding factors and further improve prediction performance. We introduce OrthoNormal basis construction In cOnfounding factor Normalization (ONION) to remove confounding covariates and use the Domain-Adversarial Neural Network (DANN) to penalize models for encoding confounder information. We apply the proposed methods to simulated and empirical patient data and show significant improvements in generalization.