Virmid: accurate detection of somatic mutations with sample impurity inference

Table 1 Accuracy and robustness of estimated α.

True α (%)	Estimated α (%)			Standard deviation (10^-2)	α range	BAF range^b
	Call based	Virmid (/LOR)^a	Virmid
1	2.64	2.56	1.61	0.19	1.23-1.90	11.11-47.50
5	6.31	5.56	4.74	0.28	4.07-5.17	11.11-46.34
10	10.3	10.50	9.86	0.51	9.34-11.03	11.43-44.12
20	20.4	19.92	19.59	0.44	18.44-20.13	11.43-38.46
30	30.4	30.33	30.28	0.48	28.48-30.79	11.11-33.33
40	39.6	40.46	39.94	0.22	39.49-40.51	11.43-28.89
50	49.2	51.38	50.72	0.23	50.46-51.31	11.11-23.53
60	56.2	61.16	60.62	0.54	59.78-61.48	10.71-19.15
70	62.4	70.28	70.05	0.16	69.82-70.44	10.00-14.29
80	67.2	80.33	80.04	0.38	79.52-80.76	9.38-10.81
90	67.4	88.91	88.88	0.53	88.06-90.00	8.82-9.68

^aEstimates without the loss of read (LOR) bias correction. The probability that a read with at least one A is mappable was set to equal the probability that a read with at least one B is mappable, that is, x(A) = x(B) (see Material and methods). Note that loss of variants (LOV) bias has already been addressed.
^bRange of minimum BAF in the observed data. The more data points Virmid observes, the smaller the minimum BAF obtained, and this is used in Virmid to correct sampling bias.

ISSN: 1474-760X