long, consistently noisy region of the read could violate the quality threshold (mean of Q9 or less across 30 bp, i.e. 87.5% mean base accuracy). The addition of HRI trimming results in a substantial improvement over our ‘analysis’ treatment, with an absolute decrease in insertion and deletion rates by 0.23% to 1%, with generally greater improvements in the insertion rate over the deletion rate ( Table 4 ). While there is asymmetry in the insertion versus deletion rate, we did not expect a disproportionate decrease in the insertion rate compared to the deletion rate after trimming. In an attempt to improve the deletion rate, we considered whether the addition of HRI 3-mers to this metric would account for our previous observation that deletions are more common in longer homopolymers, however this did not yield substantial improvements ( Table 4 ).
