[Paper Review] actifpTM: a refined confidence metric of AlphaFold2 predictions involving flexible regions

Julia K. Varga, Sergey Ovchinnikov+1 more|arXiv (Cornell University)|Dec 20, 2024

Machine Learning in Bioinformatics6 citations

TL;DR

actifpTM refines interfacial confidence by focusing on the confident portion of interactions, making it robust to flexible flanking regions and used in ColabFold.

ABSTRACT

One of the main advantages of deep learning models of protein structure, such as Alphafold2, is their ability to accurately estimate the confidence of a generated structural model, which allows us to focus on highly confident predictions.The ipTM score provides a confidence estimate of interchain contacts in protein-protein interactions. However, interactions, in particular motif-mediated interactions, often also contain regions that remain flexible upon binding. These non-interacting flanking regions are assigned low confidence values and will affect iPTM, as it considers all interchain residue pairs, and two models of the same motif-domain interaction, but differing in the length of their flanking regions, would be assigned very different values. Here we propose actifpTM (actual interface pTM), a modified ipTM measure, that focuses on the confident region of an interaction, resulting in a more robust measure of interaction confidence, even when not the full interaction is structured. actifpTM has been incorporated into ColabFold.

Motivation & Objective

• Motivate improved assessment of protein-protein interactions where flexible regions flank the interaction.
• Address limitations of ipTM when non-interacting flexible regions distort confidence.
• Propose a modified metric that concentrates on the confident interface region.
• Enhance robustness of interaction confidence across varying lengths of flanking regions.

Proposed method

• Introduce actifpTM (actual interface pTM) as a modification of ipTM.
• Focus the confidence assessment on the confident region of an interaction rather than the full interface.
• Describe how actifs considers interchain residue pairs only within the confidently interacting region.
• Discuss integration of actifpTM into ColabFold workflows.

Experimental results

Research questions

• RQ1 Can actifpTM provide a more robust confidence estimate for motif-mediated interactions with flexible flanking regions?
• RQ2 How does actifpTM compare to ipTM when the full interaction is not fully structured?
• RQ3 Does actifpTM reduce variability in confidence across different lengths of flanking regions?

Key findings

• ActifpTM yields a more robust measure of interaction confidence in the presence of flexible, non-interacting flanks.
• ActifpTM is a refined interfacial confidence metric that concentrates on the confident portion of an interaction.
• The metric has been incorporated into ColabFold, broadening access to improved confidence estimation.