Prediction of miRNA and their targets

The raw reads for small RNA sequencing were processed with the program ACGT101-miR (LC Sciences, USA) to remove adapters, junk reads, low-complexity sequences, mRNA (ftp://ftp.ensemblgenomes.org/pub/plants/release-36/fasta/hordeum_vulgare/cds/Hordeum_vulgare.Hv_IBSC_PGSB_v2.cds.all.fa.gz), repeats (V18.02; http://www.girinst.org/repbase) and common non-coding RNA families including rRNA, tRNA, snRNA and snoRNA (version 11; http://rfam.janelia.org). Subsequently, the remaining clean unique sequences with the length of 18–25 nucleotides were mapped to miRBase (version 21; ftp://mirbase.org/pub/mirbase/CURRENT/). Length variation at both 3’ and 5’ ends and two mismatches inside of the sequence were allowed in the alignment. The unique sequences mapping to barley and other plant mature miRNAs in miRBase were identified as the known miRNAs. The remained unique sequences mapping to the arm of the known precursor hairpin were considered to be new members of known miRNAs. The mapped pre-miRNAs were further compared to barley genome (http://plants.ensembl.org/Hordeumvulgare/Info/Index) to determine genomic locations. Finally, the unmapped sequences were blasted against barley genome, and then the flank 120 nt sequences were extracted to predict secondary structures using RNAfold software (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi). The criteria were adjusted as follows: (1) number of nucleotides in one bulge in stem is ≤12, (2) number of base pairs in the stem region of the predicted hairpin is ≥16, (3) the free energy for miRNA precursor should be ≤ − 15 kCal/mol, (4) length of hairpin is larger than 50 but less than 200 nt, (5) number of nucleotides in one bulge in mature region is≤4, (6) number of biased errors in one bulge in mature region should be≤2 (7) number of biased bulges in mature region is ≤2, (8) number of errors in mature region is ≤4, (9) number of base pairs in the mature region of the predicted hairpin is ≥12, (10) percent of mature in stem is ≥80. All remaining sequences meeting these parameters were considered to be potential candidate miRNAs (PC-miRNAs). The raw reads for miRNAs were normalized by the global normalization procedures [49].

The program TargetFinder was used to identify miRNA binding sites in terms of bioinformatics analysis. In addition, degradome library was constructed to further predict the target genes of miRNAs. Raw sequencing reads were filtered by Illumina’s Pipeline software (version 1.5) and then CleaveLand3.0 was used for sequencing data analysis. Finally, degradome reads were mapped to barley mRNA database. All function annotation of target genes was taken from barley CDS database in IPK (http://webblast.ipk-gatersleben.de/barley_ibsc/downloads/160517_Hv_IBSC_PGSB_r1_CDS_HighConf_REPR_annotation.fasta.gz).