Important sites for IC50-based models.

Reported are important sites, defined as sites with >0% variable importance. The Excel file contains 4 sheets, 2 for GBM and 2 for RF. For each model, one sheet shows sites in terms of HXB2 coordinates while the other sheet shows the same sites but as CATNAP alignment’s coordinates. For each bnAb, there were 5 models resulting from 5-fold cross-validation, hence the 5 columns per bnAb. Sites are 1-indexed (i.e., the first site is numbered 1). The number between parentheses is the variable importance for the site. Models were trained on sequences paired with phenotypes defined using IC50. (XLSX)本研究报道的是变量重要性占比大于0%的重要位点。该Excel文件包含4个工作表，其中2个对应梯度提升机（Gradient Boosting Machine, GBM），另外2个对应随机森林（Random Forest, RF）。针对每一类模型，一个工作表以HXB2坐标体系展示位点信息，另一个工作表则以CATNAP比对坐标系呈现同一位点的对应数据。针对每一株广谱中和抗体（broadly neutralizing antibody, bnAb），通过5折交叉验证可得到5个模型，因此每个bnAb对应5列数据。位点采用1索引编号规则（即首个位点编号为1）。括号内的数值即为该位点的变量重要性占比。所有模型均基于与表型配对的序列训练得到，表型通过IC50进行定义。（XLSX）