Development of computational pipelines for transcriptome and miRNome characterization from RNA-seq data applied to swine adipose tissue

High throughput technologies for DNA sequencing are used more and more frequently for gene expression profiling studies (RNA-seq). With respect to other techniques such as microarrays, RNA-seq has higher sensitivity in retrieving the expressed molecules and presents the advantageous feature of allowing the detection of unknown or uncharacterized transcripts. RNA-seq data processing involves several computational steps (input preprocessing for quality evaluation and cleaning; read alignment to reference genome; transcript identification, quantification, and annotation; differential expression assessment) that have to be performed in sequential order, thus resulting in a computational pipeline. Each single RNA-seq experiment can produce large amounts of data that require the use of efficient computational methods to obtain transcriptome qualitative and quantitative characterization. There are different methods that implement each conceptual pipeline step, and new ones are continuously proposed. However, because of the variety of biological questions and study designs to which RNA-seq experiments can be applied to, there is not a commonly adopted implementation of the processing workflow. In this thesis, we developed a computational pipeline for the analysis of RNA-seq data focused on the linear transcriptome, extended an existing pipeline that analyzes RNA-seq data of microRNAs (miRNAs) and miRNA-like small RNAs, and started to develop a computational pipeline for the detection and quantification of circular RNAs. The main objectives of the first two pipelines were the profiling of the set of the transcripts (transcriptome) and small RNAs (miRNome) expressed in the considered samples, by the identification of known and new RNAs. They allowed as well to investigate RNA sequence variations (such as miRNA isomiRs), transcripts and small RNAs expression levels, and to compare expression profiles between different sample groups. The pig (Sus scrofa) is a model organism for human diseases, and very important per se for the meat industry. Fat and backfat tissues are subject of very active research since fat attributes and deposition traits are in strong connection with technological aspects and quality of pig products. However, the global framework of the biological and molecular processes regulating backfat deposition in pig is still incomplete. We applied our pipelines to RNA-seq data of polyadenylated and of small RNAs from pig subcutaneous adipose tissue samples from 20 Italian Large White (ILW) individuals. Selected animals were reared under very standard conditions but presented, for fat traits, extreme and divergent phenotypes (FAT and LEAN pigs) and genetic merits. The backfat transcription profile was characterized by the expression of 23,483 genes, of which only 54.1% were represented by known genes. Of 63,418 expressed transcripts, about 80% were non-previously annotated isoforms. By comparing the expression level of FAT vs. LEAN pigs, we detected 86 robust differentially expressed transcripts, 72 more expressed in fat pigs (including ACP5, BCL2A1, CCR1, CD163, CD1A, EGR2, ENPP1, GPNMB, INHBB, LYZ, MSR1, OLR1, PIK3AP1, PLIN2, SPP1, SLC11A1, STC1) and 14 less expressed (including ADSSL1, CDO1, DNAJB1, HSPA1A, HSPA1B, HSPA2, HSPB8, IGFBP5, OLFML3). Overexpressed genes were implied particularly in immune system processes, response to stimulus, cell activation and skeletal system development. Underexpressed genes included five heat shock proteins and were involved in unfolded protein binding and stress response functional categories. Adipose tissue alterations and impaired stress response are linked to inflammation and, in turn, to adipose tissue secretory activity, similar to what is observed in human obesity. MiRNAs play important roles in cell differentiation and physiology acting as post-transcriptional regulators of gene expression by silencing targeted transcripts. The pig backfat miRNome showed the expression of hundreds of small RNAs, including putative new miRNAs, new miRNA isoforms (isomiRs), and new moRNAs, likely produced from the terminal regions of non-canonically processed hairpin precursors. From a first study on two samples, we detected 222 known miRNAs, 68 new miRNAs and 17 moRNAs expressed from known hairpins, and 312 new miRNAs expressed from 253 new hairpins. The expression of five small RNAs, including moRNA ssc-moR-21-5p and a miRNA from a new hairpin, was validated by a qRT-PCR assay, thus confirming the robustness of our results. A second study on 18 samples identified a largely overlapping miRNome in terms of expressed elements and variations, and was important to identify differentially expressed miRNAs and moRNAs in FAT and LEAN subjects. We predicted putative regulatory interactions between small RNAs and transcripts by sequence analysis, using custom target predictions on reconstructed transcript sequences and miRNA isomiRs. We then integrated target prediction results with combined analysis of miRNA and transcript expression data, to eventually select miRNA-transcript relations most supported by negative correlation of expression profiles. Further, the predicted network of miRNA-transcript interactions was enriched by information on transcript differential expression, functional annotations and coding potential predictions, and transcript overlap with pig QTL genomic regions. In this way we were able to focus on a restricted and possibly most significant number of interactions that need to be experimentally investigated. Additional considerations are coming from the study of the possible impact of specific differentially expressed miRNAs to genes belonging to the pathways most germane to adipose tissue features. The applicative results of these studies enlarged the knowledge of transcripts and small RNAs expressed in the pig adipose tissue, as well as small RNA-transcripts regulatory interactions, providing information helpful for a better understanding of ILW pig backfat and future studies on gene expression regulation in this tissue. Moreover, the methods presented here are currently undergoing further development and extension, and have applications well over and above those presented in this thesis