Pushing the limits of structure prediction in regions of disorder using ColabFold: Progress and insights

Abstract

Understanding the impact of amino acid substitutions on protein structure is critical for unraveling mechanisms of protein function and disease. ColabFold, a user-friendly implementation of AlphaFold2, offers an efficient and accessible platform for high-accuracy protein structure predictions. In this study, we explore the utility of ColabFold in predicting structural changes arising from specific amino acid substitutions. Using a structurally well-characterized model protein, mutations were selectively introduced and resulting structural deviations compared to the wild-type conformations. By employing a confidence score embedded in the program called the predicted local distance difference test (pLDDT), which is used to assess a models accuracy, together with what is known about the proteins structure, we identify mutation-induced alterations and determine their potential implications on structural integrity. The results reported here demonstrate that ColabFold is capable of capturing subtle structural perturbations, offering insights into structural stability and the propensity for specific amino acid residues to induce or disrupt secondary structure conformations. We anticipate these studies and others like it will pave the way for high-throughput mutation screening and provide a valuable tool for protein engineering along with hypothesis-driven experimental studies to facilitate our understanding of the behavior of disease-associated mutations.