Causal inference enables the estimation of outcomes of interventions from observational mass spectrometry (MS)-based proteomics experiments
Date:
We developed a new method that brings causal inference to mass spectrometry (MS)-based proteomics, allowing us to estimate the effects of interventions—like drug treatments—without having to run the experiments. By combining Bayesian modeling with biological knowledge graphs from the INDRA database, and tackling challenges like noise, missing values, and irrelevant network connections, this approach predicts how proteins respond to perturbations directly from observational data. I validated it on both simulated and real experiments, showing that it can outperform neural networks in predicting protein responses, including in studies of drug effects on transcription factor activity.