#38 · Systems biology

IsomiR-resolved modeling links isomiR targeting and extracellular vesicle provenance

The dominant thread is treating isomiRs as distinct entities rather than collapsing them into parent miRNA abundance. PARiS probabilistically reassigns technical-error isomiR reads in small RNA-seq data to likely biological miRNA sources before downstream analysis. misoTar, a deep learning framework, predicts miRNA and isomiR-mRNA target interactions, trained on 6.66M pairs from 67 human samples. A separate preprint analyzing 96 public EV and source samples reports EVs enriched for isomiRs while parent miRNAs remain correlated, and a random forest links EVs to their cellular origins. What is new is isomiR-resolved modeling applied to targeting and provenance rather than only abundance estimation. A benchmark of normalization and differential expression methods offers adjacent quantification context, while a published partial information decomposition study addresses miRNA-miRNA synergy and competition rather than isomiRs directly.