蛋白质二级结构预测方法的评价

A method for assessing methods for protein secondary structure prediction

目前评价蛋白质二级结构预测方法主要考虑预测准确率,并没有充分考虑方法自身参数对方法的影响。本文提出一种新型评价方法,将内在评价与外在评价相结合评价预测方法的优劣。以基于混合并行遗传算法的蛋白质二级结构预测方法为例,通过内在评价,合理选取内在参数——切片长度和组内类别数,有效提高预测准确率,同时,通过外在评价,与其他基于随机算法的蛋白质二级结构预测算法比较和与CASP所提供的结论比较,说明了方法的有效性与正确性,以此验证内在评价和外在评价的客观性、公正性和全面性。

In the current methods for assessing methods for protein secondary structure prediction,the prediction precision is mainly considered,but the effect of internal parameter on prediction method is not sufficiently considered.This paper presents a new evaluation method,which assesses the advantages and disadvantages of prediction method by combining internal evaluation and external evaluation.With protein secondary structure prediction based on hybrid parallel genetic algorithm for example,reasonable selection of internal parameters（slice length and structure group class number） effectively improves the pre diction precision by using internal evaluation;on the other hand,its comparison with protein secondary structure prediction based other random algorithms and results provided by CASP shows the effectiveness and correctness of prediction.This example verifies objectivity,impartiality and comprehensiveness of internal and external evaluation.