Authors
Ihor Kendiukhov
Published in
PloS one. Volume 21. Issue 8. Pages e0347366. Epub Aug 12, 2026.
Abstract
Researchers routinely take a gene-regulatory network measured in one species-for example, which transcription factor switches which gene on or off in the mouse-and assume the same wiring holds in another species after matching genes by ancestry (orthology). How often this assumption holds quantitatively has not been measured systematically. Here I test it directly. Using single-cell RNA-sequencing of matched organs, I measure, for each transcription factor, how strongly its expression tracks that of every other gene (a co-expression "edge"), compute these edges independently in human and mouse lung, and ask how well they agree. Edges agree strongly when the organ is the same in both species (rank correlation ρ=0.74; the direction of an association-activating versus repressing-matches for 100% of the strongest edges), far beyond chance (z = 112 against a permutation null). Agreement is, however, highly uneven: edges of identity-defining factors such as NKX2-1 (lung), ERG (blood vessel), and RUNX3 (lymphocyte) transfer almost perfectly (ρ>0.80), whereas edges of signal-activated factors such as HIF1A, STAT1, and CTNNB1 barely transfer at all (median ρ=0.14 for the signal-responsive group). Strikingly, changing the organ within one species disrupts edges more than changing the species within one organ: 76% of factors are better conserved across species than across tissues. The conserved signal is not merely a by-product of matching cell-type proportions: after statistically removing cell-type identity as a covariate, cross-species agreement remains substantial (ρ=0.50). Extending the comparison to zebrafish using a real developmental cell atlas (430 million years of divergence), edge agreement falls to weakly positive (ρ=0.15-0.19), and-contrary to a simple expectation-the identity-versus-signaling gap is no longer statistically distinguishable at that distance. Whole-protein sequence identity is a poor predictor of edge transfer at mammalian distances (R2 < 0.01). Finally, benchmarking against a curated database of known regulatory interactions shows that co-expression strength is a weak indicator of direct regulation (area under the ROC curve ≈0.52), so what transfers across species is conserved co-expression structure-largely cell-identity programs-rather than validated direct regulatory wiring. Together these results give practical, quantitative guidance on when cross-species transfer of regulatory information is safe and when it is not.
PMID:
42585234
Bibliographic data and abstract were imported from PubMed on 13 Aug 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 6
- Comments 0