As expected, some reads matched both species, and so were counted twice, inflating the total percent alignment above 100% (Fig. 1A, HISAT2_sep). The percent error was calculated as the sum of the differences between actual vs. observed proportion of reads for each species. The high number of non-matching reads contributed to an unexpectedly high percentage error, as shown in the “HISAT2_sep” bars in Fig. 1B. The accuracy, defined as the percentage of human reads aligned with human genome, was relatively insensitive to the species proportions (Fig. 1C). Separating the alignment produced greater total alignments with greater error and a similar accuracy, compared with mixed-genome alignment.
