Lin, J.; Liu, X.; Zhu, H.

Diverse mouse models have been generated to study human tumors. Although mouse and human tumors share similar differentially expressed genes and cancer hallmarks, many drugs that work in mice fail in humans. What makes mouse models poorly recapitulate human tumors remains unclear. We postulate that transcriptional regulation by lineage-specific long noncoding RNAs (LS lncRNAs) critically determines cross-species and cross-tumor differences. To test this hypothesis, we identified LS lncRNAs, predicted their target genes, integrated 9,058 RNA seq samples from 13 human tumors and their mouse counterparts, and analyzed transcriptional regulation by LS lncRNAs across cellular contexts. LS lncRNAs substantially and tumor specifically reconfigure transcription and signaling, and strongly influence cancer immunity and anti cancer drug efficacy. These results provide systematic information for exploring and interpreting human tumor mouse models and for identifying human and tumor-specific diagnostic and therapeutic targets. They also present an analytical approach applicable to other human diseases and their mouse models.