三种生物信息学软件对癫痫相关SCN1B错义突变的预测性能评估

Evaluation on Performance of Three Bioinformatics Software for Prediction of SCN1B Missense Mutations Associated with Epilepsy

目的使用三种生物信息学软件(MutPred2、PROVEAN和SNAP2)对癫痫相关SCN1B基因错义突变的预测性能进行评估。方法从HGMD数据库收集22个具有明确致病性的错义突变作为阳性组,从dbSNP和ExAC数据库收集136个无疾病表型报道的错义突变作为阴性组,评估三种软件的预测性能。结果从准确率、特异度、PPV、NPV、MCC值和AUC值指标进行评估,排名依次均为PROVEAN、SNAP2、MutPred2。PROVEAN软件阳性判断的阈值为-2.5,位于SCN1B错义突变阳性组和阴性组预测分值的中位数(-3.199^-2.026)之间。结论三种软件在使用不同指标参数进行评估时的性能有所不同,其中PROVEAN软件综合多个指标参数对SCN1B基因错义突变的预测性能最佳。

Objective To evaluate the prediction performance of three bioinformatics softwares(MutPred2, PROVEAN, and SNAP2 on SCN1B missense mutations associated with epilepsy. Methods From Human Gene Mutation Database, dbSNP and Ex AC database, 22 missense mutations confirmed with reported phenotype in patients with epilepsy were used as positive set, while 136 missense mutations without clinical significance were used as negative set. These selected variants were used for evaluation of three softwares. Results In terms of accuracy, specificity, positive predict value, negative predict value, MCC score and AUC score, the rank was PROVEAN, SNAP2, and MutPred2. The score cutoff(-2.5) of PROVEAN was between the median prediction scores of positive and negative sets(-3.199 to-2.026).Conclusions The prediction performance of three kinds of software may be different when using different parameters, among which PROVEAN has the best prediction performance on SCN1B missense mutations.