Putative target genes were manually selected from these candidates based on their location in the maize genome. Functions of the predicted target genes were assigned manually according to the functions of the best hits from the BLAST search [41] and [43] against the NCBI database (http://www.ncbi.nlm.nih.gov/blast/Blast.cgi). For the predicted novel miRNA sequences, conservation in other plant species was examined by searching

the nucleotide databases with BLASTn [41] to identify their homologs and surrounding sequences. These germination-related miRNAs were also aligned with the maize genome using PatScan [42]. To analyze whether the matched sequence could form a suitable hairpin, the sequences of candidate precursors were analyzed using BAY 73-4506 purchase RepeatMasker (http://www.repeatmasker.org/) to eliminate repetitive sequences. Sequences surrounding the matched sequence (100–200 nt to either side) were extracted and run through RNAfold (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi). Most targets of miRNAs in plants have one miRNA-complementary site located in the coding region and occasionally

in the 3′ or 5′ un-translated regions (UTRs) [21], [36], [38], [44] and [45], and plant miRNAs exhibit perfect or near-perfect complementarity with their target mRNAs [46]. We adopted a set of previously reported rules to predict miRNA targets [36] and [47]. These rules allow one mismatch in the region complementary to nucleotide positions 2 to 12 of the miRNA, do not allow a mismatch at position 10/11, which is a predicted cleavage site, and allow three additional mismatches between positions 12 and 22, but with no more than two continuous find more mismatches. Therefore, candidate miRNA

target genes were determined using publicly available prediction algorithms, including miRU [48], the target search in WMD web [49], and the prediction tool in the UEA plant sRNA toolkit. These programs were used with their default settings. The microarrays used in this study were obtained from GSE9386, entitled “Genome-wide analysis of gene expression profiles during the kernel development of maize (Z. mays L.)”. The raw data from microarray hybridization was exported from GenePix suites Selleck Venetoclax 6.0 (Axon, USA) and imported into LIMMA with annotation and spot types [50]. Spots with a negative flag value were assigned a weighting of 0.1 in the subsequent analysis. Background-subtracted signal intensities were normalized using two-step normalization, consisting of print-tip group loss (within-array normalization) and between-array scale normalization. The adjusted p value was then assessed using the false discovery rate. To identify a statistically significant differential expression of genes, p = 0.01 was used as a criterion. To obtain probe annotations, the consensus representative sequences of all probes were searched using BLAST against the TIGR rice protein database (http://www.tigr.