Predicting Protein–protein Association Rates using Coarse-grained Simulation and Machine Learning

• DOI: 10.1038/srep46622
• Publication Date: 2017-04-18T10:08:44Z
• Authors: Zhong-Ru Xie; Jiawen Chen; Yinghao Wu
• Keywords: Benchmark (surveying); Association (psychology); Training set

Abstract Protein–protein interactions dominate all major biological processes in living cells. We have developed a new Monte Carlo-based simulation algorithm to study the kinetic process of protein association. tested our method on previously used large benchmark set 49 complexes. The predicted rate was overestimated test compared experimental results for group hypothesized that this resulted from molecular flexibility at interface regions interacting proteins. After applying machine learning with input variables accounted both conformational and energetic factor binding, we successfully identified most complexes association rates improved final prediction by using cross-validation test. This then applied independent similar accuracy obtained training set. It has been thought diffusion-limited is dominated long-range interactions. Our provide strong evidence also plays an important role regulating studies insights into mechanism offer computationally efficient tool predicting its rate.