Identification of Key Residues in SFN0011 Protein Structure Through Network Centrality Analysis and Computational Machine Learning Approaches.

Hamid Latifi-Navid1 , Zahra-Soheila Soheili1 *, Shahram Samiei2 , Mehdi Sadeghi3

  1. Department of Molecular Medicine, National Institute of Genetic Engineering and Biotechnology, Tehran, Iran
  2. Blood Transfusion Research Center, High Institute for Research and Education in Transfusion Medicine, Tehran, Iran
  3. Department of Medical Genetics, National Institute for Genetic Engineering and Biotechnology, Tehran, Iran

Abstract: Protein function depends on the composition and distribution of amino acid residues. Furthermore, the identification of key residues in the structure of protein-ligand complexes provides good targets for mutagenesis experiments for functional optimization and development of novel methods for designing next generation drugs. Residue interaction networks (RINs) analysis is performed on 3D protein structures to determine key residues. The identified residues may be recognized in surface or core, alone or in protein complexes. Moreover, it is well established that the driving forces for protein complex formation are not equally distributed across the contact surfaces. Indeed, Hot-Spots (HS) residues are a small set of amino acids that play crucial roles in protein to protein binding. Here, we used network centrality analyses and computational machine learning approach on SFN0011-ligand complexes and identified key residues in their structures and interactions.

Methods: To perform residue and betweenness centrality analyses (RCA and BCA), predicting protein dynamics and determine key residues in contact surfaces, Chimera, Cytoscape, structureViz2, RINalyzer, RINspector, DynaMine, and SpotOn tools were recruited. RCA examines the average shortest path length change under removal of individual nodes. However, BCA evaluates how often a node is crossed by paths between all node pairs. Nodes with a Z-score greater than 2, represent a set of central residues. Also, SpotOn provides a robust algorithm with a demonstrated sensitivity of 0.98 and accuracy of 0.95 on an independent test set. HS residues are those which, generate a binding free energy difference (ΔΔG binding) ≥2.0 kcal/mol.

Results: We determined 107, 129, and 14 key residues through BCA, RCA, and Machine Learning approaches. The result of the DynaMine server revealed three regions; rigid (925 residues), context-dependent (601 residues) and flexible (325 residues) in SFN0011-ligands complex structure.

Conclusion: Integration of RCA, BCA results revealed 70 common residues in SFN0011-ligands complex structure. Distribution of the residues in three highlighted areas was as follows: rigid (4 residues), context-dependent (6 residues), flexible (60 residues). Amongst 14 Hot-Spots (HS) residues, four amino acids were distinguished between two major chains (A,B) of SFN0011 protein and ten amino acids were identified at the protein-ligand contact surfaces.





اخبــار



برگزار کنندگان کنگره


حامیان کنگره