Biomacromolecular Quantitative Structure–activity Relationship (bioqsar): a Proof-of-concept Study on the Modeling, Prediction and Interpretation of Protein–protein Binding Affinity
Peng Zhou,Congcong Wang,Feifei Tian,Yanrong Ren,Chao Yang,Jian Huang
DOI: https://doi.org/10.1007/s10822-012-9625-3
IF: 4.179
2013-01-01
Journal of Computer-Aided Molecular Design
Abstract:Quantitative structure–activity relationship (QSAR), a regression modeling methodology that establishes statistical correlation between structure feature and apparent behavior for a series of congeneric molecules quantitatively, has been widely used to evaluate the activity, toxicity and property of various small-molecule compounds such as drugs, toxicants and surfactants. However, it is surprising to see that such useful technique has only very limited applications to biomacromolecules, albeit the solved 3D atom-resolution structures of proteins, nucleic acids and their complexes have accumulated rapidly in past decades. Here, we present a proof-of-concept paradigm for the modeling, prediction and interpretation of the binding affinity of 144 sequence-nonredundant, structure-available and affinity-known protein complexes (Kastritis et al. Protein Sci 20:482–491, 2011) using a biomacromolecular QSAR (BioQSAR) scheme. We demonstrate that the modeling performance and predictive power of BioQSAR are comparable to or even better than that of traditional knowledge-based strategies, mechanism-type methods and empirical scoring algorithms, while BioQSAR possesses certain additional features compared to the traditional methods, such as adaptability, interpretability, deep-validation and high-efficiency. The BioQSAR scheme could be readily modified to infer the biological behavior and functions of other biomacromolecules, if their X-ray crystal structures, NMR conformation assemblies or computationally modeled structures are available.