Additional file 3: of Mechanistic interpretation of non-coding variants for discovering transcriptional regulators of drug response

Transcription factor binding motifs. We obtained 239 PWMs from three different data sources: (1) factor book motifs processed by ENCODE UCSC; (2) motifs downloaded from Factorbook; (3) HOCOMOCO Human (v10) motifs. STAP models are trained separately for each motif. We excluded a motif if the maximum prediction of the corresponding STAP model is too small (< 0.5). The remaining 225 PWMs, which were used in further analysis, are included in Additional file 3. (TXT 72 kb)

转录因子结合基序数据集。我们从三类不同数据源共获取239个位置权重矩阵（Position Weight Matrix，PWMs）：其一为经ENCODE UCSC处理的Factorbook基序；其二为从Factorbook下载的基序；其三为HOCOMOCO人类（v10版）基序。针对每个基序单独训练STAP模型，若对应STAP模型的最大预测值过小（小于0.5），则剔除该基序。最终剩余的225个位置权重矩阵将用于后续分析，已纳入补充材料3（TXT格式，72 KB）。