Fig. 6

Steps 1 and 2 of the workflow displayed in Fig. 1. Step 1, “Short reads align long reads,” leads to alignments of the long read to be corrected with short reads. Given the alignments, putative variant positions P1:X, P2:X, and P3:X are identified (“X” indicates mismatch). Step 2, “Phase and filter out reads,” inspects the sequence content of both the long and the short reads at each of these positions. Different scenarios are conceivable. In P1:X, all short reads agree on “A,” so no short reads are removed because of this position. In P2:X, there are large amounts of short reads for supporting both “A,” agreeing with the long read, and “C,” disagreeing with the long read. Because the number of short reads in disagreement with the long read is too large, these short reads are removed (“filtered out”); most likely, they stem from a different haplotype. We now refer to this position as P2:M with “M” meaning “match,” indicating that there is nothing further to be done. In P3:X, all but one short read support “G.” Because the amount of short reads (exactly one) that do not support “G” is too small, these short reads are kept; most likely, they contain a sequencing error. After these two steps, P1:X and P3:X require further attention, while P2:M has been resolved