Beschreibung:
Abstract: In recent years, the drug discovery paradigm has shifted toward compounds that covalently modify disease-associated nucleophilic proteins, because they tend to possess high potency, selectivity, and duration of action. The rational design of novel targeted covalent inhibitors (TCIs) typically starts from resolved macromolecular structures of nucleophilic proteins in their apo or holo forms. However, the existing TCI databases contain only a paucity of covalent protein–ligand (cP–L) complexes. In this respect, CovPDB was developed, the first database dedicated to high-resolution cocrystal structures of biologically relevant cP–L complexes, curated from the Protein Data Bank. For these curated complexes, the chemical structures and warheads of pre-reactive electrophilic ligands, as well as the covalent bonding mechanisms to their target nucleophilic proteins, were expertly manually annotated. Totally, CovPDB contains 733 proteins and 1,501 ligands, relating to 2,294 cP–L complexes, 93 reactive warheads, 14 targetable residues, and 21 covalent mechanisms. In addition, the application of the “quantum chemical electrophilicity index” (ω) for covalent warhead reactivity was investigated. In this work, a method was established that applies the concept of electrophilicity index to estimate the reactivity potential of covalent compounds targeting cysteine. For leadlike molecules (molecular weight > 250 Da), a truncation algorithm was applied before reactivity calculations. Whereas for compounds with molecular weight (MW) below 250 Da, the electrophilicity index was directly used to estimate compound reactivity. A total of 148 covalent compounds with 8 different warheads were analyzed. This analysis demonstrated that the electrophilicity index calculated for compounds containing vinyl carbonyl, aldehyde, nitrile, and ketone warheads showed no correlation with binding affinity in our dataset. This suggested that while electrophilicity is an important factor in determining the reactivity of covalent warheads, it may not be the direct factor affecting binding affinity. Among the 14 targetable residues in CovPDB database, the cysteine side chain has a free thiol group, making it the amino acid residue most often covalently modified by TCIs, thereby prolonging on-target residence time and reducing the risk of idiosyncratic drug toxicity. Hence, to identify targetable cysteines, a novel ensemble stacked machine learning (ML) model to predict hyper-reactive druggable cysteines, called HyperCys was developed. It is anticipated that HyperCys will be an effective tool for discovering new potential reactive cysteines in a wide range of nucleophilic proteins and will provide an essential contribution to designing TCIs with high potency and selectivity