Skip to main content
. 2021 Jun 7;12:3370. doi: 10.1038/s41467-021-23544-8

Fig. 1. Different combinations of signal and noise parameters result in indistinguishable observed distributions of AI values.

Fig. 1

a Two simulated parametrizations (left) of true AI signal (AItrue; solid line) and noise (dashed line) that combine to produce overlapping observed AI values (right; red and blue, respectively). These and similar observations are indistinguishable by Mann–Whitney–Wilcoxon and Kolmogorov–Smirnov tests; see Supplementary Note S1. AI distributions are shown for allelic coverage 100. Noise distribution shown at AItrue = 0.5. Both signal and noise are modeled using beta-binomial distributions; the following parameters are shown: [ρSignal = 0.001, ρNoise = 0.1] and [ρSignal = 0.1, ρNoise = 0.001]; simulation sample size 500,000. Other coverage levels and combinations of ρ values are shown in Supplementary Note 1. b Quantile–Quantile (QQ)-plot for distributions set by parametrizations 1 and 2 from panel (a). Quantiles were taken from 0 to 1, with step 0.01.