Datasets: Benchmark datasets from Deng et al. 2022 (Nucleic Acids Research, doi:10.1093/nar/gkac112). FN = rRNA input (measures sensitivity); FP = non-rRNA input (measures specificity); FN+FP = mixed.
| Dataset | Test | Pairs | Description |
|---|---|---|---|
| SILVA_rRNA | FN | 20,000,000 | SILVA SSU+LSU rRNA sequences |
| OMA_CDS | FP | 20,000,000 | Prokaryotic and eukaryotic mRNA |
| oma_silva | FP | 1,027,675 | OMA mRNA CDSs with ≥70% identity to rRNA genes; FPR on rRNA-similar mRNA |
| homd_fp | FP | 100,558 | HOMD oral-microbe mRNA CDSs with ≥70% identity to high-FPR rRNA hits in OMA_CDS |
| ENA_virus | FP | 27,206,792 | Viral gene sequences from ENA |
| Amplicon_16S | FN | 7,917,920 | Real 16S V1-V2 amplicon reads (oral microbiome) |
| Human_ncRNA | FP | 6,330,381 | Human non-coding RNA |
| MetaT | FN+FP | 9,165,829 | Oral metatranscriptome: 4.7M prokaryotic mRNA, 2.5M human mRNA, 73K viral mRNA, 1.9M rRNA (21% rRNA fraction) |
Metrics: Sensitivity = (total - misclassifications) / total for FN datasets. FPR = misclassifications / total for FP datasets. MetaT reports reads classified as rRNA vs. the expected ~21% rRNA fraction.
| Dataset | Type | Total pairs | Misclassifications / rRNA classified | Metric | Value | Wall time (s) | Memory (MB) |
|---|---|---|---|---|---|---|---|
| oma_silva | nonrrna | 1027675 | 15966 | FPR | 1.55% | 3628 | 3805 |
| homd_fp | nonrrna | 100558 | 1604 | FPR | 1.60% | 15 | 3784 |
| OMA_CDS | nonrrna | 20000000 | 1448 | FPR | 0.01% | 9103 | 3814 |
| SILVA_rRNA | rrna | 20000000 | 5345 | Sensitivity | 99.97% | 4139 | 4358 |
| Amplicon_16S | rrna | 7917920 | 7 | Sensitivity | 100.00% | 1271 | 4214 |
| Human_ncRNA | nonrrna | 6330381 | 2003 | FPR | 0.03% | 141 | 3798 |
| MetaT | mixed | 9165829 | 1888130 (20.6% of reads classified as rRNA; ~21% expected) | NA | NA | 994 | 3981 |
| ENA_virus | nonrrna | 27206792 | 217 | FPR | 0.00% | 367 | 3792 |