Table 1.
Performance Comparison of Mosaic CNV Detection Tools
| Algorithm | Install | Runtime | Sensitivity | Specificity | URL |
|---|---|---|---|---|---|
| MONTAGE | Easy | Short (35 s/10sa) | Good(1/1) | Good(0/0) | https://github.com/CAG-CNV/MONTAGE |
| MoChA | Difficult | Long (1m1sb) | Good(1/1) | Good(0/0) | https://github.com/freeseek/mocha |
| RGADA-MAD | Difficult | Short (14 s) | Low(0/1) | Low(1/0) | https://github.com/isglobal-brge/MAD |
| BAFSegmentation | Easy | Long (1m14s) | Good(1/1) | Low(186/0) | http://baseplugins.thep.lu.se/wiki/se.lu.onk. BAFsegmentation |
| triPOD | Easy | Very Long (10 m) | Low(0/1) | Low(0/0) | https://github.com/jdbaugher/tripod |
Install ease based on actual setup with non-superuser credentials, not exclusively the documented setup instructions provided by the algorithm. Runtime listed per sample 610 k density SNP microarray. aSorted by chromosome and position input file. bEagle phasing pipeline (1 m) and Chromosomal alterations pipeline (1 s) steps included. Sensitivity and Specificity based on running the same sample data through each algorithm and comparing results. In parenthesis is Observed / Expected mosaic CNV calls. See Fig. 7 for additional Sensitivity/Specificity analysis where we demonstrate in 755 samples a 0.975 false positive rate 0.344 (MONTAGE) vs. sensitivity of 0.920 at false positive rate 0.598 (MoCha) vs. sensitivity of 0.280 at false positive rate 0.627 (RGADA-MAD)