Using Biological Networks to Guide Biomedical Prediction

Predictive modeling is a transformative tool for understanding complex biological systems and advancing biomedical discovery. A central challenge is ensuring that predictive models are not only accurate but also biologically interpretable. One way to address this challenge is through network-guided approaches, which integrate prior biological knowledge into bioinformatic algorithms and model architectures. By aligning predictive models with biological networks at various scales, these approaches can improve biological interpretability while maintaining strong predictive performance. In this chapter, we introduce the fundamentals of biological networks and describe strategies for incorporating networks into modeling frameworks for predicting biomedically relevant phenotypes. These concepts are explored via a case study of network-guided drug response prediction, in which hierarchical knowledge of cell biology is instrumental to achieving interpretable predictions and biological insights into chemoresistance.