Fig. 3

Sensitivity and false discovery trade-off curves of aggregation methods. Twenty simulated experiments based on parameters estimated from biological data were analyzed with different aggregation methods and averaged producing (a) and zoomed in (b). sleuth in gene mode (“sleuth-Gene”) is a standard gene-level differential analysis method. Aggregation results based on transcript p values are shown using two approaches: sleuth transcript p values aggregated by the Lancaster method (“sleuth-Lancaster Tx”) and sleuth transcript p values aggregated by the Šidák-adjusted minimum method (“sleuth – Sidak Tx”). Finally, sleuth TCC p values obtained by running sleuth on TCC counts were aggregated with the Lancaster method (“sleuth-Lancaster TCC”). Dashed lines indicate true FDR at 0.01, 0.05, and 0.1. The shapes (circle, triangle, square) on each sensitivity-FDR curve indicate the true FDR and sensitivity at each method’s reported FDRs of 0.01, 0.05, and 0.1