Rational Prediction of PROTAC-Compatible Protein–Protein Interfaces by Molecular Docking

This article has been Reviewed by the following groups

Read the full article See related articles

Listed in

Log in to save this article

Abstract

No abstract available

Article activity feed

  1. future PROTAC-based drug design projects willuse AF2 predicted structures

    For AlphaFold structures to be useful in such a workflow, two considerations come to mind. First, the distinction between proximity-based activity and the need for specific and tight interactions, as mentioned in our comment above. If proximity alone is important, then even somewhat inaccurate AlphaFold structures could be quite useful. On the other hand, if specific interactions are important, then would the methods presented here be able to account for inaccuracies in, for example, surface-exposed loops and side-chain orientations? Second, if protein dynamics are highly relevant for PROTAC-induced PPIs, would the static structures predicted by AlphaFold be appropriate inputs for docking-based prediction methods? The authors might consider including a more extended discussion on the relevance of inferred protein structures to the methods developed in this work.

  2. The dataset is composed of three types of E3 ligases(9 VHL17,54–59, 2 Cereblon60–63 and 2 cIAP64)

    The docking methods presented here may be suitable for selecting appropriate E3 ligases for a certain target. We suggest screening multiple E3 ligases against the same protein of interest to determine if a given E3 can be prioritized for PROTAC development at this early stage in the drug development pipeline. Would the computational pipeline presented be able to suggest a particular PROTAC-POI pair?

  3. Introduction

    This paper focuses on specific protein-protein interactions induced by PROTACs. There appears to be some evidence that specific PPIs promote PROTAC efficacy, but it would be nice if the authors could summarize the evidence for proximity alone versus specific, PROTAC-induced binding interactions (i.e. interactions stable enough for crystallization). Is it always the case that a specific protein-protein interaction leads to higher PROTAC activity? Or is proximity sufficient? Or just the overall orientation of important residues in the E3 ligase and the target protein? A discussion of this point would help motivate the focus on crystal structure interactions.

  4. Abstract

    This review was prepared at Arcadia Science by Jasmine Neal and Jase Gehring.

    The authors present a novel workflow for predicting protein-protein interactions in a ternary complex between a PROTAC, a target protein of interest, and a E3 ligase. They propose a computational workflow involving LightDock, energy rescoring with VoroMQA, and finally filtering by distance between user-defined residues on the two proteins. The authors used 13 known crystal structures as validated interactions to evaluate their approach. While this approach could recapitulate known binding interaction modes, the overall success rate was rather low, and experimentally determined interactions were not always highly ranked. This approach does outperform AlphaFold-Multimer, which was unable to capture known PPIs, and performs more or less on par with the more computationally expensive method PROTAC-Model. The manuscript is clear and well-written, and the authors present their results without exaggerating the effectiveness of their methods. Overall, PROTACs have attracted huge research interest, and this paper highlights the current challenges in predicting protein-protein interactions promoted by PROTACs. Although the methods presented here did not result in high confidence predictions, the methods do recapitulate known binding modes, especially when constrained by user-selected residues. This work this work serves as an impactful advance for future PROTAC protein-protein interaction prediction capabilities