4.6. In Silico miRNA Target Prediction and Overrepresentation Analysis of Reactome Pathways in Predicted Target mRNAs

To facilitate use of the previously generated miRNA target prediction resource for Atlantic salmon, MicroSalmon (https://github.com/AndreassenLab/MicroSalmon/ (accessed 20 September 2022)) [23], the corresponding full-length mRNAs (Genbank Accession {"type":"entrez-nucleotide","attrs":{"text":"GIYK00000000","term_id":"2045700002","term_text":"GIYK00000000"}}GIYK00000000) [22] of each DE-mRNA from the microarray analysis was identified in the following manner. The Oligo sequences of the DE-mRNAs were aligned to the FL transcriptome using the blastn tool in the BLAST+ package (v.2.9.0) [87], and FL sequences with at least one oligo sequence passing a filter of 90% sequence identity and 91% query coverage were classified as FL DE-mRNAs. The FL-DE-mRNAs and the DE-miRNAs from each of the two materials were used in conjunction with the MicroSalmon database to identify mRNA transcripts which were both differentially expressed and predicted to be targets of differentially expressed miRNAs, which were in turn were classified as FL-targets of a particular DE-miRNA (tDE-mRNA). The oligo IDs associated with each tDE-mRNA were processed in the STARS pipeline to identify overrepresented TM and TS categories in the subgroup of tDE-mRNAs.

The complete FL sequences, the predicted gene symbols and BLAST-derived natural language descriptions of the FL-targets from the transcriptome resource (Supplementary Tables S3 and S4) [22] were used as search input for the NCBI Gene database as well as the Universal Protein Resource UniProt (https://www.ncbi.nlm.nih.gov/gene (accessed 20 September 2022) and https://www.uniprot.org/ (accessed 20 September 2022), respectively), to identify gene symbols from human homologs. The non-redundant list of gene symbols of tDE-mRNAs were subsequently used as input in the PANTHER Overrepresentation Test (version 17.0 Released 12 February 2022) [88] (http://pantherdb.org/ (accessed 20 September 2022)) to identify gene pathways overrepresented in the tDE-mRNA dataset. The following parameters were used in the PANTHER Overrepresentation Test: Homo Sapiens REFLIST (containing 20,589 genes), and the Reactome version 65 Released 1 October 2021 annotation, and Fisher’s Extract and Calculate False Discovery Rate, retaining hits with an FDR < 0.05 and Fold Enrichment ≥2. The pathways were ordered hierarchically and the most specific pathway in each tree was retained. The exception to this was if the most specific pathway had the same number of representative genes as the next more general pathway above it in the hierarchy, in which case that one was retained instead. Finally, the specific pathways were grouped by manually assigned functional categories that were based on the pathways’ descriptions in the Reactome Pathway Browser (https://reactome.org (accessed 20 September 2022)) and the associated literature. The overrepresentation analysis results were plotted with ggplot using a publicly available R-script [89] to generate Figure 3 and Figure 4.