Table 2.
For two of the transmark benchmarks (transmark00-all which has no simulated frameshifts and transmark02 which has a 2% indel rate) we calculated average alignment coverage for a set of shared true positives (the set of test sequences correctly annotated by all tools). For transmark00-all the set includes 10,582 hits and for transmark02 the set includes 6,620 hits. For each sequence in these sets, we compared the true embedding coordinates of the test sequence to the alignment coordinates reported by each tool. To calculate alignment coverage, we record the percentage of all the test sequence nucleotides that are included in a true positive alignment. If there is more than one true positive alignment per tool, we combine the coverage of those alignments, being sure to only count each nucleotide once, to get the maximum coverage across all alignments.
Average Alignment Coverage on Shared True Positives | |||||||
---|---|---|---|---|---|---|---|
Benchmark | BATH | BATH --nofs | tblastn | MMseqs2 | nhmmer | LAST | DIAMOND |
transmark00 | 95.2% | 95.2% | 93.9% | 94.9% | 81.3% | 88.6% | 94.7% |
transmark02 | 91.7% | 63.5% | 78.3% | 61.9% | 79.9% | 85.6% | 89.8% |