一个基于知识的蛋白质结构预测评分函数

评分函数设计是预测蛋白质结构的关键之一。本研究采用pab_select 25列表中的蛋白质结构作为学习样本，并按照结构测定方法的不同将其分成X-Team和NMR—Team两组。根据学习样本中两两氨基酸残基的距离分布，分别求出两组评分函数。然后使用公认的测试样本数据评估上述评分函数的性能。结果表明：两组评分函数在识别结果上存在一致性，且总体性能较高，但X-Team评分函数的识别质量高于NMR—Team评分函数。

A Knowledge Based Scoring Function for Protein Structure Prediction

Designing a scoring function plays a key role in protein structure prediction.In this paper,the protein structures,which were determined with experimental methods,from pdb-select 25 were employed as training dataset for designing a scoring function.The training data were divided into two sub-datasets,X-Team and NMR-Team respectively,according to the methods for determining their structures.Based on the pair-wise residues' distance distribution of the training sub-datasets,two sets of parameters for scoring function were determined.The Decoys 'R' Us was employed as test dataset to validate the designed scoring function.It was proved that both of the scoring functions gained from X-team and NMR-team respectively were highly effective while the former was better.