Exact-Sequence Profiling Reveals Small RNA Fragment Remodeling in Hepatocellular Carcinoma Across Dog, Human, and Mouse Datasets

Background/Objectives: Hepatocellular carcinoma (HCC) is associated with altered microRNA (miRNA) expression, but small RNA sequencing (sRNA-seq) analyses usually aggregate reads by annotated features. This aggregation can mask changes in individual miRNA sequence variants (isomiRs) and in fragments derived from transfer RNAs (tRNAs), ribosomal RNAs (rRNAs), and other non-coding RNAs. We aimed to characterize HCC-associated small RNA changes at exact-sequence resolution. Methods: We analyzed five liver sRNA-seq datasets from dogs, humans, and mice. The main canine, human (hsa-Tissue 1) and mouse (mmu-Tissue 1) comparisons included only three biological samples per group. Each unique read sequence was retained as the analytical unit and annotated afterwards against species-specific RNAcentral references. We compared sequence-level and feature-level differential expression, characterized fragment terminal positions, and examined sequence sharing and fold-change concordance between datasets. Results: In canine HCC, 3520 of 40,231 retained sequences were differentially abundant. Sequence-level analysis resolved HCC-associated isomiRs and tRNA-derived fragments that were not apparent in feature-level summaries, and additional changes involved rRNA-derived and unannotated sequences. Several RNA classes showed recurrent terminal patterns compatible with selective processing or stabilization. Exact sequences shared between datasets showed dataset-dependent concordance of HCC-associated changes, including limited agreement between the two human datasets. Conclusions: Exact-sequence analysis complements feature-level quantification by distinguishing HCC-associated changes among individual small RNA sequences and provides candidate fragments for functional studies. The small sample sizes and differences in disease context, control tissue, and library preparation limit the interpretation of individual candidates and cross-dataset comparisons.