The local sequence-specific context (high GC, low complexity) can lead to lower base quality or alignment success, that is, systematic noise. For each base sequenced, we used the data from the four patient blood samples to estimate its systematic noise. The significance of the noise was measured with a binomial statistics using the local error rate (fraction of non-reference bases) calculated from 20 adjacent positions (10 upstream, 10 downstream). The P-value was Bonferroni-corrected for multiple testing in a single 200-bp amplicon (P < 5 × 10-4). Positions declared recurrently noisy in three or more control blood samples were removed from the list of candidate variants.
