The Human Targetable Protein Networks In The Ligand Cluster- And Sequence-based Space

Biologists traditionally classify proteins based on their sequences, structures and functions. Here, we present a complementary approach to quantitatively distinguish proteins at the level of ligand cluster. We extracted 237,960 protein-ligand interactions from ChEMBL. By the 2D structure similarity,13,769 ligand clusters emerged from 158,087 distinct ligands which were recognized by 1,501 proteins. We presume that if two proteins interact with more common ligand clusters, they are more similar at the level of ligand cluster. This approach was compared with the approach by global sequence similarity, and two kinds of similarity were mapped to the ligand cluster- and sequence-based networks (LCBN and SBN). The protein communities emerged from both networks were then compared with the gene families proposed by HGNC (Human Genome Organization Gene Nomenclature Committee). We found although there were some disagreements between the LCBN and SBN, communities in both networks were largely the same with normalized mutual information at 0.9. In the end, we exemplified how to interpret the meaning of the protein module in the LCBN with biological and chemical information. From the perspective of ligand cluster, we could utilize or avoid the multiple pharmacological effects that result from the chemical crosstalk between proteins. Last, we integrated the targets, ligands, ligand clusters, pathways, diseases and drug adverse reactions into our database, ePlatton, which could help users to query the possible target range of their interested molecules and exam the promiscuity of these molecules.