Table 13.
The Relative Sensitivity and Specificity of BLAT at Various Settings With and Without a Best-Match Filter
| All nonhuman mRNA alignments | Near perfect | CPU days | % genome | % RefSeq | Enrichment | |
|---|---|---|---|---|---|---|
| K | N | |||||
| 5 | 1 | no | 8.69 | 0.90% | 69% | 77x |
| 4 | 2 | no | 6.06 | 0.83% | 68% | 81x |
| 8 | 1 | yes | 95.25 | 0.86% | 69% | 80x |
| Only the best alignment for each mRNA | Near perfect | CPU days | % Genome | % RefSeq | Enrichment | |
| K | N | |||||
| 5 | 1 | no | 8.69 | 0.70% | 67% | 95x |
| 4 | 2 | no | 6.06 | 0.66% | 65% | 99x |
| 8 | I | yes | 95.25 | 0.68% | 66% | 97x |
The human genome (August 2001 freeze UCSC assembly http://genome.ucsc.edu) was aligned against a collection of 86,000 nonhuman mRNA sequences totaling 123,000,000 bases taken from Genbank. The CPU days were measured on 800 MhZ Pentiums. The percentage genome column shows what percent of bases in the human genome are part of gapless alignments with the mRNAs. The percentage RefSeq shows the percentage of coding bases in human RefSeq-defined genes (from the database at http://genome.ucsc.edu) which are part of a gapless alignment. Enrichment is the ratio of the RefSeq coding density inside of the alignments compared to coding density in the genome as a whole. Higher levels of enrichment indicate greater specificity. Also shown are the same statistics when only the single best place that a nonhuman mRNA aligns was considered.