Scoring Function And Molecular Automatic Optimization Program Based On Protein Pocket Residues

The biological activity of molecules is often one of the most important indicators of early drug discovery,and a large number of biological activity experiments require huge economic and time costs.Therefore,the prediction of affinity through computers has always been a research hotspot.In order to fully consider the synergy of various interaction energies of different protein amino acid residues,we have developed a new machine learning based scoring function(rbScore).The scoring function is trained by using random forests and 107 descriptors which obtained by dispersing the various interactions of the ligand and the protein pocket to the main and side chain of 20 different amino acid residues.Here we use the PDBbind 2017 refined-set and 2013 core-set datasets to train and test our scoring function model respectively.The Pearson correlation coefficient(Rp)of the test results is 0.79,and the standard deviation(SD)is 1.50.In addition,molecular optimization is also an important process in drug design.Bio-isosteric replacement is a common method for molecular structure optimization in medicinal chemistry,and it has a wide range of applications in the range for improving selectivity,activity,ADMET,and evading patent.However,finding a suitable replacement structure is not a simple matter,and the use of computational methods can help pharmaceutical chemists find suitable bio-area replacement structures.With reference to the principle of protein-based isosteric substitution,we have developed a molecular automated optimization program based on protein pocket local environment.The program is mainly divided into two parts.On the one hand,We processed and analyzed a large number of protein-ligand complex three-dimensional structure data,generated a fragment-sub-pocket environment pair,and constructed a feature fingerprint and stored it in a database to provide a queryable bio-electronic isostere.;On the other hand,we provide automated programs that include molecular cutting,combined sub-pocket fingerprint generation,combined sub-pocket fingerprinting,similarity alignment and more.Through the operation of the program,the specified segments can be automatically replaced by biological isosters to obtain new molecules for reference.