De-extinction: what the biology allows

A species is not a list of genes, and that single fact organises everything de-extinction can and cannot do. The workflow — read the extinct genome, edit a living relative, gestate the embryo in a surrogate — is real at every step, and each step loses information.

What ancient DNA can and cannot give

After death, DNA is attacked by water and by microbes: cytosine deaminates, which reads as a C-to-T substitution, and the strands break into short fragments — tens of bases in anything warm; permafrost and caves preserve longest. Recovered genomes are therefore low-coverage mosaics assembled against the reference genome of a living relative, so whatever diverged — repetitive regions, regulatory sequence, unrepresented loci — is either invisible or silently inferred. Comparison with the surrogate yields candidate trait loci, not a complete specification: partly the extinct genome is a reconstruction whose gaps were filled with the nearest living relative’s pieces.

Editing a proxy is not rebuilding a species

Differences between species are mostly polygenic and regulatory: thousands of small-effect variants, expression timing, developmental programs — much of it in non-coding DNA whose function is unknown. Multiplex editing can install a bounded set of named traits — coat character, cold tolerance, variants associated with body size — into a living relative, and the result is exactly that: an edited relative expressing selected derived traits, not the extinct animal. What editing cannot reach is everything not specified in sequence you possess: the epigenetic setting of embryogenesis, gut and skin microbiomes assembled horizontally, and learned behaviour. Social species transmit migration routes, calls and predator recognition culturally; an animal raised without conspecific teachers is ecologically naive even if its genome were complete — and the teachers are, by construction, extinct.

Gestation is the biological bottleneck

The embryo needs a womb, and placentation is species-specific: implantation signalling, gestation length, foetal-maternal immunology and sheer size mismatch all reject distant surrogates. The credible routes run through close relatives — nuclear transfer into related oocytes, or stem-cell-derived embryos carried by a related host — and each costs years of reproductive physiology per species, which no amount of sequencing accelerates. Rearing then repeats the problem outside the womb.

Where the honest part lies

What already works is genetic rescue of the living: cryobanked material from an endangered population carries alleles the survivors have lost, and cloning or assisted reproduction can put them back — a bounded, checkable intervention with no behavioural fiction attached. De-extinction as practised today sits at the proxy end of that same spectrum, and the wording matters: what returns to the reserve is an edited surrogate carrying reconstructed traits, and whether it can occupy the extinct species’ ecological role is a field experiment, not a laboratory result.