pGReMLIN: prediction of ligands using conserved substructures in protein-protein interaction

One of the sources of data for the protein dataset being explored in bioinformatics is the Protein Data Bank (PDB). Searching for similar conserved structures in biological problems was an early attempt and is still a popular method widely used by researchers. In this work, a graph-based approach named pGReMLIN is proposed, which predicts similar structures in protein-protein or protein-peptide complexes using the established conserved structural arrangements of Serine protease and BCL-2. The protein-protein or protein-peptide complexes are downloaded from the PDB, and the interface between proteins at the atomic level is modeled as a graph, with atoms as nodes and the non-covalent interactions between atoms as edges. The computation of a similar pattern of structural arrangement in the data graph was achieved using a set of candidate pairs of data graph D(G) and query graph Q(G) based on state space representation and feasibility rules that embed equality of structure and attributes. pGReMLIN was able to compute, in the data graph, a conserved structural arrangement that represents highly conserved interactions at the specificity binding interface of trypsin and trypsin-like proteins. Also, our method was able to identify similar patterns with large pattern size in the data graph common to Serine protease and BCL-2 and as well as unique to Serine protease and BCL-2. Furthermore, pGReMLIN was able to present various large patterns of similar conserved patterns that can be further explored to determine their potency. Hence, the presence of similar patterns in our protein-protein dataset indicates the possibility that patterns found in our dataset have similar functions to the conserved structures. Keywords: Protein-protein interaction. Graph-subgraph isomorphism. Ligands prediction.