Identification of known and novel miRNAs in B. rapa
We aligned the clean reads with known miRNAs in the miRBase (https://www.mirbase.org/), allowing two mismatches using TopHat (http://ccb.jhu.edu/software/tophat/index.shtml). The potential miRNAs were mapped to B. rapa reference genome (http://brassicadb.cn/#/) to extract the 600-bp flanking sequence. The flanking sequence of miRNAs were also aligned with known pre-miRNAs to confirm their identities of miRNA family. The flanking sequence of unknown small RNAs was also used to predict secondary structures using RNAfold software (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi). To identify novel miRNA, Small RNAs that showed less than 2 mismatches with known miRNAs in miRbase were filtered. The potential miRNAs that had fewer than 10 hits in the reference genome were retained. 600-bp flanking sequences of small RNAs were used to predict their secondary structures using RNAfold software with default parameters. Novel miRNAs were identified according to the plant miRNA criteria (Axtell et al., 2018, Yu et al., 2011). In brief, a key judgement was made by determining whether the potential miRNAs and their miRNA* were on the stem with fewer than four bulges, and the miRNA candidate were most abundant among small RNAs generated from given miRNA precursors. After Blast these novel miRNA precursors in NCBI, we found them were only detected in Brassica, therefore, they were defined as Brassica-specific miRNAs.