论文部分内容阅读
目的将ROC曲线分析应用在计算机虚拟药物筛选研究领域中。方法利用分子对接软件Surflex以β-分泌酶的三维晶体结构为靶,虚拟筛选50个抑制剂和9950个无活性化合物,利用5种打分函数(Surflex-score,D-score,G-score,PMF-score,ChemScore)对筛选结果打分排序,应用ROC曲线分析打分排序后的结果。结果ROC曲线分析出利用D-score打分函数得出的曲线下的面积(AUC)最大,为0.935;如果特异度作为优先考虑因素时,应用Surflex-score打分函数可以获得最多的活性化合物;如果灵敏度作为优先考虑因素时,应用D-score打分函数可以获得最多的活性化合物。结论ROC曲线分析同时考虑到了灵敏度和特异度,在计算机虚拟药物筛选过程中,可以根据实际情况设定打分函数的阈值,获得尽可能多的活性化合物。
Objective To apply ROC curve analysis in the field of computer virtual drug screening research. Methods Surflex-score (D-score, G-score, PMF) was used to detect the three-dimensional crystal structure of β-secretase using molecular docking software Surflex. Fifty inhibitors and 9950 inactive compounds were screened. scoring, ChemScore) to sort the screening results, using ROC curve analysis to sort the sorted results. Results The ROC curves showed that the area under the curve (AUC) obtained by the D-score scoring function was the largest, 0.935. If the specificity was the highest priority, the Surflex-score scoring function could be used to obtain the most active compounds. If the sensitivity As a top priority, the most active compounds can be obtained using the D-score scoring function. Conclusions The ROC curve analysis takes into account both sensitivity and specificity. In the computer virtual drug screening process, the threshold value of the scoring function can be set according to the actual situation to obtain as many active compounds as possible.