Protein Sequence Predicted by Using Parallel CRF Method Based on Backbone Angle
Combining advance mathematic model to predict protein structure is one of the most challenging problems in structural biology. Condition Random Fields(CRF) is shown a powerful algorithm by many examples of informatics and widely used in protein structure predicted. CRFsampler can automatically optimizes more than ten thousand parameters quantifying the relationship among primary sequence and backbone angle; In this paper, we construct a parallel CRF protein sequence predicted model; by using backbone structure, the Cb is set up(GLY is pseudo), dihedral torsion angles are calculated. Between sequence and backbone angles, the parameters of feature is found by optimizing. The residue predicting accurate rate is 24.07％, the GLY predicting rate high to 64％. The rate is over 25％ in the case of SAS>75％. The rate is also high when contact number small or lager.
component protein CRF sequence prediction parallel computation
Shaoping Chen Xing Wang Shesheng Zhang Jim Zhang
Department of Mathematics Wuhan University of Technology Wuhan 430070, China Wuhan shi, Jiang An Qu Er Qi Lu 145 Hao Wuhan 430012, China
国际会议
香港
英文
211-213
2010-08-12（万方平台首次上网日期，不代表论文的发表时间）