Identifying Perturbation Targets Through Causal Differential Networks

Learn to identify perturbation targets in biological systems through causal differential networks in this conference talk by Rachel Wu from Valence Labs. Discover a novel causality-inspired approach that addresses the practical challenges of drug target discovery and cell engineering when working with high-dimensional biological data containing thousands of variables but limited samples per intervention. Explore how this method infers noisy causal graphs from both observational and interventional datasets, then maps the differences between these graphs along with statistical features to identify intervention targets. Understand the joint training framework that combines both modules using simulated and real biological data, and examine how this approach consistently outperforms baseline methods across seven single-cell transcriptomics datasets. Gain insights into the significant improvements this method demonstrates over current causal discovery approaches for predicting both soft and hard intervention targets across various synthetic datasets, making it particularly valuable for applications where biological systems don't adhere to classical causality assumptions.