Global Optimization by Conformational Space Annealing and its Applications to Protein Structure Prediction/Determination and Machine Learning

Research Seminar
Title: Global Optimization by Conformational Space Annealing and its Applications to Protein Structure Prediction/Determination and Machine Learning
Speaker: Prof. Jooyoung Lee
Director of Center for In Silico Protein Science
Professor, School of Computational Sciences
Korea Institute for Advanced Study
Time: 13:00-14:30, Mar. 9, 2018
Location: Youcai Deng Hall，School of Life Sciences
Host：Xiaodong Su
Abstract：
First, I will discuss our recent progresses on the protein structure prediction using the methodology of global optimization as illustrated in the CASP11/12 competitions held in 2014/2016. We will demonstrate that this method can be applied to difficult MR (molecular replacement) targets to determine X-ray crystallography structures of proteins and protein complexes, which could not be solved using conventional MR methods. We will also discuss the possible application of our method to the high throughput NMR structure determination of large proteins (over 20 kDa) and membrane proteins.First, I will discuss our recent progresses on the protein structure prediction using the global optimization method of Conformational Space Annealing (CSA) as illustrated in the CASP11/12 competitions held in 2014/2016. We will demonstrate that this method can be applied to difficult MR (molecular replacement) targets to determine X-ray crystallography structures of proteins and protein complexes, which could not be solved using conventional MR methods. We will also discuss the possible application of our method to the high throughput NMR structure determination of large proteins (over 20 kDa) and membrane proteins.
If time is allowed, I will also discuss the optimization issue in the study of machine learning (ML). A preliminary study indicates that proper application of CSA to ML can provide a solution to the overtraining problem in ML. I will share the progress of our attempt to build our own AlphaGo in this respect.
Welcome！