Performance on single sample clinical data:
To test the performance of ROHMM on single sample clinical data
we used the publicly available case data from Pippucci and colleagues
(Pippucci et al., 2013). This test was also a means to show the unique
abilities of ROHMM in a single sample case where homozygous
reference sites are almost always not available from the variant call
format. We tested ROHMM using 3 different settings to infer
homozygosity from this data. ROHMM was able to detect the long
homozygous stretch containing the CACNA2D2 NM_006030.4:c.1295del
mutation in the proband as it was detected by the original study
(Pippucci et al., 2013). Authors of the original study used a predefined
set of SNPs to infer homozygosity and we tried simulating a similar
input using a BED file containing the same set of SNPs withROHMM ’s spike-in function (Setting3). We observed that the
spike-in functionality further enabled the detection of potentially
cryptic short ROH’s that are otherwise not visible from the variant
sites only data (Figure 7).