Computational protocol: Computational analysis of calculated physicochemical and ADMET properties of protein protein interaction inhibitors

Similar protocols

Protocol publication

[…] The iPPIs dataset (compounds with bioactivity below 30 μM) was built by merging all compounds from IPPI-DB minus bromodomain’s inhibitors while adding 89 compounds extracted from the TIMBAL database targeting menin-mixed lineage leukemia (MLL) and neuropilin, and 24 small molecule disruptors of the glucokinase–glucokinase regulatory protein interactions. Regarding the “non-iPPIs” modulators, all datasets were created using the version 14 of the ChEMBL database categorization which is available at ftp.ebi.ac.uk/pub/databases/chembl/Allosterism. We extracted these molecules from the version 20 of the ChEMBL database with the highest ChEMBL confidence score of 9 and a bioactivity below 30 μM. We formed one category of allosteric molecules (allosteric modulators of kinases, proteases, phosphodiesterases, phosphatases, nuclear receptors, ion channels and GPCRs) and four categories of orthosteric molecules (nuclear receptors, ion channels, GPCRs and enzymes (proteases, kinases, phosphodiesterases and phosphatases)). From the same database, we retrieved the non-allosterics oral bioavailable approved drugs (OMD) and we extracted from it the natural product-derived compounds (NPD). We removed from both these subsets putative allosterics modulators showing occurrences in the Allosterics ASD database. Then, all datasets were treated following the same filtering and diversity search protocols. We performed with the FAF-Drugs3 web-server the selection of compounds within the 150 to 900 Da area (filtered subset). On these molecules, we applied a clustering protocol with the Accelrys Pipeline Pilot FCFP4 fingerprints (maximum Tanimoto coefficient of 0.2) where the centroid of each cluster was taken to build the diversity subset. In order to have a relatively similar number of chemically diverse compounds in each dataset, we kept the entire diversity subset when its amount was below 650 compounds, otherwise we proceeded a random picking after diversity searching (random subset). To insure that diversity or random subsets represent properly the original filtered subsets, we visualized the chemical space of each subset using the path-based fingerprints projection visualization tool of StarDrop 6.1. In the same space we visualized the filtered subsets (light blue), the diversity subsets (red) and the random subsets (white) (see ). [...] We used the FAF-Drugs3 web-server to compute physicochemical descriptors: number of rotatable bonds, rigid bonds, HBAs, HBDs, rings, charges (formal charges at pH 7), heavy atoms, carbon atoms, heteroatoms and stereocenters, MW, log P, log D (at pH 7), TPSA, maximum size of ring, number of rings and aromatic rings, flexibility, total charge and Fsp3 . We also derived the Lipinski’s RO5, the Pfizer’s 3/75 rule and the Golden Triangle. The estimation of the chemical beauty was carried out with StarDrop v6.1 while the water solubility was computed with the pkCSM server. […]

Pipeline specifications

Software tools FAF-Drugs, pkCSM
Application Drug design