We rotated though each algorithm to determine the calls of the verification set. For a given algorithm’s verification set calls, we tested the evaluation set calls of every algorithm. We used this approach rather than a consensus-based method, as we did not want to favor or disfavor any particular algorithm or group of algorithms. Sensitivity was calculated as in the simulation benchmark, now with true differential expression defined by an adjusted P value <0.1 in the larger verification set, as diagrammed in Additional file 1: Figure S18. Figure 8 displays the estimates of sensitivity for each algorithm pair.
