Results of the different analysis methods on the HCV data. (a) Upper plot: Venn-Diagram of the hits at the gene level using the CELL-BASED, AVERAGE and RIPLEY analysis methods. A total of 30 host dependency factors were identified using RIPLEY, 44 host dependency factors were identified using the AVERAGE, and 88 factors were identified using the CELL-BASED approach. 39 genes were commonly identified between CELL-BASED and AVERAGE, but only 8 genes common between AVERAGE and RIPLEY and only 10 common genes between RIPLEY and CELL-BASED. Lower plot: Venn-Diagram of the hits at the gene level using the CELL-BASED, GSEA-ONLY and MARS-ONLY analysis methods (b) Receiver operator characteristic (ROC) analysis of correct identification of positive and negative controls in the HCV screen. Sensitivity and specificity of the recognition of positive and negative controls was computed for different thresholds on computed scores or significance levels, using the AVERAGE, CELL-BASED and RIPLEY approaches. Hit thresholds were computed on z-score, clustering scores and ES, respectively. ROC curves were generated by varying these thresholds, and plotting sensitivity over 1-specificity. Pink: CELL-BASED, Black: AVERAGE, Red: RIPLEY. The area under each curve was computed to obtain a single value measuring the quality of the control classification. AUC values of 0.5 correspond to random guessing, AUC values of 1 to perfect classification. Achieved AUC values varied between 0.87 for RIPLEY, 0.95 for AVERAGE, and 0.99 for CELL-BASED, showing best performance of the CELL-BASED approach. (c) Comparison of obtained area under the ROC curve (AUC) values on an HCV validation screen, using hit genes identified in the primary screen using the CELL-BASED and RIPLEY methods. In brief, hit genes identified using the AVERAGE method on the primary screen were subjected to a secondary validation screen. The intersection of predicted hits on the primary screen using the RIPLEY and CELL-BASED approaches with genes screened in the validation screen was used to compute ROC curves and AUC values for CELL-BASED and RIPLEY. Shown are AUC-values over different z-score thresholds on the validation screen.