Unveiling Antibody-Mediated Allostery in Interleukin-1β Via Conformational Sampling and Machine Learning

Abstract

Human interleukin-1β (IL-1β), a pivotal proinflammatory cytokine, is a therapeutic target in autoimmune and inflammatory diseases. While antibodies blocking IL-1β signaling are effective, their allosteric mechanisms remain poorly understood. This study investigates how four distinct antibodies induce long-range allosteric effects in IL-1β, leveraging computational approaches to map allosteric communication and identify critical sites. Ensembles of apo and antibody-bound IL-1β states were generated using the enhanced conformational sampling technique ClustENMD, followed by the application of two different dimensionality reduction methods in machine learning (principal component analysis, PCA; linear discriminant analysis, LDA) to the generated conformers. PCA highlighted how diverse ensembles ClustENMD generated, while LDA revealed antibody-specific allosteric effects on the human IL-1β. By integrating conformational dynamics with machine learning, this work advances a predictive framework for engineering antibodies with tailored allosteric properties. The discovery of binding sites on IL-1β might further open avenues for drug design.