Aligner | Simulated Reads | Real Reads |
---|
 |
Execution time (hrs)
|
TPR
|
PPV
|
Execution time (hrs)
|
% reads uniquely mapped
|
BSMAP | 32.1 | 81.5 | 99.5 | 20 | 72.2 |
mrsFAST | 20.8 | 77.5 | 83.9 | 23.8 | 72.8 |
Pash high | 12.4 | 80.5 | 99.5 | 13.5 | 84.4 |
Pash fast | 6.1 | 80.2 | 99.1 | 5.8 | 81.6 |
RMAP-BS | 2.7 | 82.7 | 99.7 | 2.5 | 73.2 |
- Each dataset used for evaluation contains 1 million reads. Execution time, in hours is reported for the simulated dataset. We report the true positive rate (TPR), defined as the overall percentage of correctly mapped reads relative to all the input reads, and the positive predictive value (PPV), which indicates what percent of all reported mappings are correct. TPR is a measure of sensitivity, and PPV is a measure of specificity. Execution time, in hours, and percent of reads unambiguously mapped are reported for the real trimmed WGS bisulfite-treated reads. Pash 3.0 is run with two sensitivity settings: high and fast. Pash is 1.5-5.2x faster than BSMAP, 1.6-4.1x faster than mrsFAST, and 2.32x slower than RMAP-BS. On the simulated dataset BSMAP and RMAP-BS are more sensitive, but Pash maps 8-11% more real trimmed reads at unique locations compared to the other aligners.