Soroush Akbari‐Ardabili, Safiyeh Aghazadeh
Rational, mechanism-based drug selection for canine diffuse large B-cell lymphoma (DLBCL) remains limited, despite extensive molecular work in human DLBCL. We asked whether human and canine DLBCL share a conserved transcriptional network that could anchor therapeutic hypotheses for dogs. Using public human and canine DLBCL transcriptomes, we derived a 409-gene cross-species module defined by concordant tumour-normal dysregulation and projected it onto pathway and protein-protein interaction (PPI) networks. The module was dominated by DNA replication, cell-cycle checkpoint and DNA damage-response programmes and contained a compact PPI core centred on TOP2A and PARP1. Querying the Connectivity Map/LINCS library with the human component of the module highlighted anthracycline-like topoisomerase II poisons (valrubicin, etoposide, amsacrine) and the PARP inhibitor rucaparib among the ~0.2% most negative connectivity scores, while directly targeting TOP2A and/or PARP1. A simple up-down module score strongly separated tumour from normal lymph nodes in both species and was associated with outcome: higher scores predicted longer progression-free survival in CHOP-treated dogs and better overall survival in an independent human cohort, with a similar but underpowered trend in TCGA-DLBC. Finally, extracellular-vesicle microRNA (EV-miRNA) profiling in human DLBCL showed that differentially expressed EV-miRNAs, including let-7 family members, miR-21-5p, miR-124-3p and miR-205-5p, converge on the same TOP2A/PARP1-centred core. These cross-species, multi-layer data support topoisomerase II poisons and PARP inhibition as coherent, network-anchored candidate therapies for canine DLBCL, with module scores and EV-miRNAs as candidate biomarkers.