Phylostratigraphy and the evolutionary origin of cancer genes

Domazet-Loso and Tautz (2010) described phylostratigraphy as a way to match the origin of founder genes with particular macroevolutionary transitions.¹ It rests on a model of genome evolution in which complex phenotypic innovations are accompanied by the emergence of founder genes, whose descendants can still be traced in living organisms.¹ They treated the origin of multicellularity as such a transition, one that would have needed new gene functions.¹ Because cancer can be viewed as a malfunction of interaction between cells, they argued that it is tightly tied to multicellular life.²

To trace human cancer genes, Domazet-Loso and Tautz built a database of genome sequences divided into 19 phylostrata, matched to the relationships of the major taxa.³ The test they set out was simple. If the origin of cancer genes carried a signal tied to the origin of cancer, such genes should be overrepresented in one evolutionary period on the phylostratigraphic map.⁴ They proposed that tracking cancer genes in this way should also give insight into the origin of multicellularity.²

Levin and colleagues later used the same strata to read the transcriptomes of engineered cell collectives. Pai and colleagues compared basal Xenobots with controls. The two sets were equally enriched until the Vertebrata stratum, but Xenobot transcripts were more enriched in the older Bilateria, Eumetazoa and Metazoa strata.⁵ Gumuskaya and colleagues reported that anthrobots shift toward more ancient gene expression.⁶ They also counted genes from the two oldest strata to see how far anthrobots behaved like unicellular organisms.⁷ Fotowat and colleagues found that most upregulated genes in neurobots were among the most ancient, unlike biobots and shams.⁸

Pio-Lopez and Levin (2026) took gene ages for 19,660 human protein-coding genes from Litman and Stein and noted that phylostratigraphy has limits that affect the accuracy and consistency of its findings.⁹ Ortholog identification must be consistent across databases. Litman and Stein found a consensus across several databases, but uncertainty may remain for some genes, giving false positives and false negatives.⁹ Noncoding genes are harder to analyse, because a much lower fraction of them can be assigned a reliable age.⁹