Order of designs by predicted binding score.

Numbers of correctly ranked design mutation pairs with large affinity difference. All mutation pairs for which there is an affinity difference of at least 50-fold are investigated. All design pairs with these mutations (i.e. for each of these pairs there are as many design pairs as scaffold crystal structures) are checked, if the order of the mutations by total score or binding score is the same order as by affinity. A cell shows the number of correctly ordered design pairs, and the number of all design pairs. The mean for this part is calculated by scaling the percentage of a test case by the number of mutation pairs (i.e. NOT by design pairs, which would bias the value too much towards test cases with many crystal structures).

本统计针对亲和性差异显著的设计突变对（design mutation pairs）的正确排序情况展开。研究选取所有亲和性差异至少达50倍的突变对（mutation pairs）作为分析对象。针对携带上述突变的全部设计对（design pairs），即每一组突变对所对应的设计对数量与支架晶体结构（scaffold crystal structures）的数量一致，我们将验证：基于总得分（total score）或结合得分（binding score）排序的突变顺序，是否与基于亲和性的排序结果一致。每个单元格将分别展示正确排序的设计对数量，以及总设计对数量。本部分的平均值计算方式为：以单个测试用例（test case）的百分比乘以该用例的突变对数量，而非设计对数量——若采用设计对数量计算，会使平均值偏向于晶体结构数量较多的测试用例，产生统计偏倚。