Unexpected similarity between HIV-1 reverse transcriptase and tumor necrosis factor revealed by binding site image processing

Abstract

Rationalizing the identification of hidden similarities across the repertoire of druggable protein cavities remains a major hurdle to a true proteome-wide structure-based discovery of novel drug candidates. We recently described a new computational approach (ProCare), inspired by numerical image processing, to identify local similarities in fragment-based subpockets. During the validation of the method, we unexpectedly identified a possible similarity in the binding pockets of two unrelated targets, human tumor necrosis factor alpha (TNF-α) and HIV-1 reverse transcriptase (HIV-1 RT). Microscale thermophoresis experiments confirmed the ProCare prediction as two of the three tested and FDA-approved HIV-1 RT inhibitors indeed bind to soluble human TNF-α trimer. Interestingly, the herein disclosed similarity could be revealed neither by state-of-the-art binding sites comparison methods nor by ligand-based pairwise similarity searches, suggesting that the point cloud registration approach implemented in ProCare, is uniquely suited to identify local and unobvious similarities among totally unrelated targets.AUTHOR SUMMARYComputational comparison of binding sites in proteins can provide insights on potential unrelated proteins that may bind to similar ligands. However, accurate prediction of binding site similarity requires powerful methods, ideally able to detect even local similarities. We herewith applied a recently developed computer vision method to identify an unexpected binding site similarity between two totally unrelated proteins that was confirmed experimentally by in vitro biophysical binding assays. Considering more precisely local similarities can therefore efficiently guide drug discovery, notably to repurpose existing drug candidates.

Subpocket Similarity-Based Hit Identification for Challenging Targets: Application to the WDR Domain of LRRK2

“…107,828 subpockets, extracted from the sc-PDB database of druggable protein−ligand complexes, 53 were aligned to the LRRK2-WDR cavity using a point cloud registration algorithm (ProCare), 10 specifically designed to find the best possible alignments between two clouds of cavity points and estimate their local similarity. 11,12 Given the high proportion of polar features in the query cavity, alignment by the color descriptor (c-FH) was chosen to prioritize subpockets in three steps (Figure 2): (i) selection of 2,852 subpockets with a ProCare similarity score above 0.47, a threshold previously shown to optimally discriminate known similar from known dissimilar pocket pairs, 10 and merging of the corresponding fragments in the target cavity coordinate frame using the same rotation/translation matrix as that selected for aligning their subpockets; (ii) prioritization of 389 fragments (derived from noncofactors) for which at least 50% of their pharmacophoric features match, within a upper distance of 3.0 Å, a LRRK2-WDR cavity point of compatible pharmacophoric property; (iii) sampling and adding fragments from sc-PDB bound cofactors with respect to the observed cofactor-bound targets enrichment, yielding a final set of 412 fragments. The fragments were located along three delimited zones from the upper side to the lower side of the inner core (Figure 2).…”

“…In the case of partial similarity (just a part of the query cavity is similar to a part of a remote pocket), transferring ligands from one pocket to another one is much more difficult and at best usually restricted to shared fragments. 10,11 To address this difficult issue, we recently proposed a novel computational approach (POEM: Pocket Oriented Elaboration of Molecules) 12 that generates ligands from the simple knowledge of a target cavity, by automatically linking fragments originally bound to remote proteins exhibiting only local similarities to the query cavity. When applied to a protein kinase (CDK8), POEM was able to quickly generate a single-digit nanomolar inhibitor in two Design-Make-Test-Analyze (DMTA) cycles 13 and 43 synthesized ligands.…”