Table 3.
Wall clock time (s) and maximum RAM usage (GB) for algorithms on assigning simulated short reads (SR)
| Algorithm | SR to REF-1 | SR to REF-2 | ||
|---|---|---|---|---|
| Time (s) | RAM (GB) | Time (s) | RAM (GB) | |
| Mora | 477.15 | 47.24 | 657.80 | 47.52 |
| AugPatho2 | 28588.09 | 51.58 | 29278.85 | 15.99 |
| Pufferfish | 101.64 | 10.89 | 325.29 | 10.54 |
| Bowtie2 | 726.44 | 2.91 | 894.84 | 26.85 |
| Kraken2 | 73.48 | 2.40 | 70.54 | 24.33 |
| Clark* | 300.97 | 45.69 | 272.01 | 41.17 |
Other than using more RAM on short reads to REF-2, Mora consistently used fewer resources than AugPatho2 and was substantially faster. Time cannot be directly compared between the Mora/AugPatho2 and Kraken2/Clark as they require the output from one of the initial aligners. Clark has an asterisk to signify that had errors in building the REF-2 database, which resulted in less resources used. CPU time and other information can be found in Additional file 1: appendix Table A.4. All algorithms were run with 8 threads on a Google cloud virtual machine instance of type c2-standard-60