Ancient genes can be served as pan-cancer diagnostic and prognostic biomarkers.

One important challenge for cancer is efficient biomarkers monitoring its formation and developments remain greatly limited. Although the accumulated big omics data provide great opportunities to the above purpose, the biomarkers identified by the data-driven strategy often do not work well in new datasets, which is one of the main bottlenecks limiting their utilities. Given that atavistic phenotype is generally observed in cancer cells, we have been suggested that the activity of progenitor genes in tumour could serve as an efficient cancer biomarker. For doing so, we first curated 77 progenitor genes and then proposed a quantitative score to evaluate cancer progenitorness. After applying progenitorness score to ~ 22 000 samples, 33 types of cancers from 81 datasets, this method generally performs well in the diagnosis, prognosis and therapy monitoring of cancers. This study proposed a potential pan-cancer biomarker and revealed a significant role of atavism in the formation and development of cancers.