Table 1.

Cutoffs for individual tools at different global confidence levels

Comp. strategy	High	Medium	Low
BLAST	95% id 95% length	92.5% id 92.5% length	90% id 90% length
Infernal	BLAST E=1e-3	BLAST E=0.1	No BLAST filter
	Infernal E=1e-3	Infernal E=1e-3	Infernal E=1e-3
Infernal miRNA	Not applied	BLAST E=1e-3	BLAST E=0.1
		Infernal E=1e-9	Infernal E=1e-6
tRNAscan-SE	High	High	As is
RNAmmer	As is	As is	As is
snoStrip	As is	As is	As is
miRDeep	Hand cleaned	As is	As is

The results from the individual tools are merged at 3 different cutoff levels: high, medium and low. This table shows the correspondence between the cutoff levels of the merged annotation and those of the tools. For each computational (comp) strategy, we define these levels. Note that Infernal screen have been divided into Infernal (without miRNAs) and Infernal miRNA which is only for miRNA families. The Infernal results were filtered by the family specific gathering scores as well. For As is, we refer to the programs default values for the respective versions (see the Methods section for version numbers) without subsequent cleaning.