Table 1.
Benchmarking of different denoising algorithms using the MOCK1 dataset
| Basic trimming | Average length | Error | CPU cost | Number of seq |
|---|---|---|---|---|
| Denoiser | 504 | 0.0045 | 112 hr | 862279 |
| Acacia | 482 | 0.0040 | 8.8hr | 845513 |
| Pre-cluster | 482 | 0.0028 | 8 hr | 845513 |
| AmpliconNoise | 499 | 0.0019 | 370 hr | 860273 |
| NoDe | 481 | 0.0012 | 9.5 hr | 845513 |
| Strict trimming | Average length | Error | CPU cost | Number of seq |
| Denoiser | 439 | 0.0024 | 96 hr | 785115 |
| Acacia | 424 | 0.0021 | 7.7hr | 827123 |
| Pre-cluster | 424 | 0.0014 | 7 hr | 827123 |
| AmpliconNoise | 424 | 0.0013 | 312 hr | 818421 |
| NoDe | 425 | 0.0008 | 8.3 hr | 827123 |
The comparison covers the final error rate as well as the computational cost (on a single CPU - Intel Xeon E5-2640 2.50 GHz) for the analysis pipelines including all tested denoising algorithms (Acacia, Denoiser, Pre-cluster, AmpliconNoise, NoDe). Also the number of reads and average read length returned by the different algorithms is displayed.