Results

A summary of the overall sensitivity, selectivity, and search speed for the four tRNA search programs tested is shown in Table 2.1. The number of true positives is based on the percentage of tRNAs detected within a test set taken from the Sprinzl tRNA database (see Table 2.2). The false positive rate is based on analysis of randomly generated sequence data (Table 2.4). The search speeds for the various programs are shown for a scan of the current C. elegans genomic sequences averaging 30 kilobases per clone. tRNAscan 1.3 search speed decreases approximately linearly with length. Search speed for tRNAscan-SE is approximately constant, but varies based on tRNA density within the sequence.

Table 2.1: Overall detection rates of tRNA search programs.

True positives are based on detection rates within a non-organellar, non-viral subset of the Sprinzl tRNA database (see Table 2.2). False positive rates are estimates based on searches of randomly generated human sequence (see Table 2.4). Search speeds are from a search of 58.4 Mbp of C. elegans cosmid sequences on a Silicon Graphics Indigo2 R4400 200 Mhz workstation.

Search Method	True Positives	False Positives	Search Speed
	(%)	(per Mbp)	(bp/sec)
tRNAscan 1.3	95.1	0.37	400
EufindtRNA	88.8	0.23	373,000
tRNA covariance model search	99.8	< 0.002	20
tRNAscan-SE	99.5	< 0.00007	30,000