Figure 5. Sensitivity and FDR for simulated dataset.

GROM-RD had the highest sensitivity and lowest FDR for duplications. GROM-RD’s sensitivity was lower than RDXplorer’s sensitivity for deletions, but GROM-RD had a much lower FDR. Ten thousand deletions and duplications were simulated from human reference hg19 using RSVSim. CNVs were biased to repeat regions. One hundred-base paired-end Illumina reads with 500 base inserts were simulated at 11x coverage using pIRS and mapped to hg19 using BWA.