MID sub-group generating

Raw reads were split into MID groups according to their 12 nucleotide barcodes. For each MID group, quality threshold clustering was used to cluster similar reads. This process groups reads derived from a common template RNA molecule together, whereas separating reads derived from distinct RNA molecules. A Levenshtein distance of 15% of the read length was used as the threshold. This was calibrated using RNA controls with known sequences (Supplementary Fig. ¹). For each sub-group, a consensus sequence was built based on the average nucleotide at each position, weighted by the quality score. In the case that there were only two reads in an MID sub-group, we only considered them useful reads if both were identical. Each MID sub-group is equivalent to an RNA molecule. Next, we merged all of the identical consensus to form unique consensus sequences, or unique RNA molecules, which were used to estimate the diversity and assess the sequencing depth in rarefaction analysis (Fig. 1c, d and Supplementary Fig. ⁴). Scripts for this section can be downloaded at https://github.com/utjianglab/MIDCIRS.