CAVIAR: a method for automatic cavity detection, description and decomposition into subcavities

Abstract

The accurate description of protein binding sites is essential to the determination of similarity and the application of machine learning methods to relate the binding sites to observed functions. This work describes CAVIAR, a new open source tool for generating descriptors for binding sites, using protein structures in PDB and mmCIF format as well as trajectory frames from molecular dynamics simulations as input. The applicability of CAVIAR descriptors is showcased by computing machine learning predictions of binding site ligandability. The method can also automatically assign subcavities, even in the absence of a bound ligand.The defined subpockets mimic the empirical definitions used in medicinal chemistry projects. It is shown that the experimental binding affinity scales relatively well with the number of subcavities filled by the ligand, with compounds binding to more than three subcavities having nanomolar or better affinities to the target. The CAVIAR descriptors and methods can be used in any machine learning-based investigations of problems involving binding sites, from protein engineering to hit identification. The full software code is available on GitHub and a conda package is hosted on Anaconda cloud. File list (2) download file view on ChemRxiv Marchandetal_MainText.pdf (2.03 MiB) download file view on ChemRxiv Marchandetal_SI.pdf (513.53 KiB)

SHREC 2022: Protein-ligand binding site recognition

“…Given a high LC score, a low PC score is related to a pocket which is exceedingly large with respect to the ligand binding region. Therefore, this suggests a margin for improvement in the direction of a higher segmentation of the returned sites into separate smaller pockets or sub-units (sub-pockets [9,17]). Given the effectiveness of the ML approaches proposed, such a segmentation might be able to identify more precisely the exact binding site without compromising unreasonably the effectiveness of the ranking.…”

“…Despite its simplicity, one of the most widely adopted and successful scoring systems is to use volume ranking [8]. Some algorithms such as grid based methods, also consider algorithm-specific descriptors (such as the degree of buriedness [9]) or chemical information based on the degree of evolutionary conservation of each residue [10,11].…”

Estimating the Similarity between Protein Pockets

“…At first glance, all geometric methods aim at identifying sufficiently buried zones unoccupied by protein atoms, but they differ in their strategies to search for these areas. Grid-based methods place the protein into a cartesian grid and identify grid cells likely to be in a cleft by analyzing their neighborhood [ 13 , 14 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 ]. POCKET [ 13 ] and LIGSITE [ 12 ], two of the earliest methods, keep cells that correspond to a ‘protein-solvent-protein’ event by scanning in three and seven directions, respectively.…”

Table1.DOCX

“…The metalloreductase activity of STEAP2 occurs across its transmembrane domains but these are neither binding sites nor druggable regions of the protein from binding site analysis using ProBiS, PrankWeb, and CAVIAR (Konc and Janežič, 2010b;Jendele et al, 2019;Marchand et al, 2021;Rocha et al, 2021). These results demonstrate that the cytoplasmic NADP-binding domain is a potential binding site but it being intracellular reduces its druggability.…”

“…All other parameters were left as default and the job was submitted for analysis at both submission webpages. The druggability of the protein was assessed using CAVIAR in the Linux terminal to identify potential cavities for ligand binding based on scores and size and ligandability among other criteria (Marchand et al, 2021). The protein chain A PDB file was used as the input file.…”