The literature, compressed

Concepts you have to have met

This field has a handful of ideas that everything else is built on, and a couple of live disagreements that are not going to be settled by the time you write your paper. If you are going to make a de novo claim, these are the ones a reviewer will already have in mind — and the first is the one that has sunk more claims than all the others combined.

1 · Homology detection failure

The whole method rests on a negative result: we searched, and found nothing outside this clade. But a search failing to find a homolog has two possible causes, and only one of them is interesting. The methodological question is whether you can put a number on the boring one. You can.

evolutionary distance from the focal species → expected bitscore genus family order class BLAST: no significant hit outside the family A · no homolog exists — the gene is new B · a homolog exists and the search could not see it byte-identical output long, slow gene detection threshold below: nothing is reported its horizon short, fast-evolving gene its horizon absence here is uninformative abSENSE — Weisman, Murray & Eddy 2020 P(search fails | homolog present, evolving normally) — many LSGs explained, but not all Synteny — Vakirlis, Carvunis & McLysaght 2020 locate the ancestral region directly; divergence is not the main source of orphans in yeast, fly, human Neither settles your gene. Both say “some but not all”. They disagree on the fraction, in different taxa, by different methods — which is a healthy disagreement, not a scandal. What settles a specific gene is its syntenic region in an outgroup. A search result is where the question starts. It is never where it ends.
The practical consequence

Report the detectability of your candidate, not just its absence. If a gene is 66 residues long and fast-evolving, say so and say what that implies for the search — because a reviewer who has read Weisman et al. will ask, and “we found no BLAST hits” is not an answer. The bitscore-versus-length walkthrough shows what this looks like on a real 66-residue protein.

2 · Proto-genes and the continuum

Carvunis et al. (Nature, 2012) looked at Saccharomyces cerevisiae and found hundreds of species-specific non-genic transcripts that are differentially regulated under stress and engaged by ribosomes. Their argument: rather than a rare birth event, there is a reservoir of proto-genes occupying a continuum from non-genic sequence to gene, and de novo birth out of that reservoir may be more common than sporadic duplication.

This reframed the question. If the continuum is real, asking when a gene was born is like asking when a dune became a hill — and the interesting measurements are about the distribution, not about individual origin stories.

3 · Preadaptation — and why it argues with the continuum

Wilson, Foy, Neme and Masel (Nature Ecology & Evolution, 2017) proposed a competing shape. The continuum hypothesis predicts young genes should be intermediate — partway to being gene-like. The preadaptation hypothesis predicts the opposite: that the de novo genes which survive are the ones that were, by luck, already unlikely to cause harm, so young genes should show extreme values of the relevant traits rather than middling ones.

They tested it on intrinsic structural disorder and found that young genes had the highest disorder of all — consistent with preadaptation, not with a gradual continuum.

Why this matters for your candidate

It is the reason a low predicted pLDDT on a young candidate is weak evidence. Both models predict young de novo proteins are disordered; they disagree about the shape of the distribution, not about the direction. Disorder is consistent with youth and does not demonstrate it — plenty of ancient proteins are disordered too.

4 · The case that teaches the most

Antifreeze glycoprotein in polar fish is the cleanest worked example in the field, because it happened twice by different routes and the comparison is the lesson.

Antarctic notothenioids — not de novo

Their AFGP gene is derived from a pancreatic trypsinogen gene. It is a spectacular new function, and it is built from an old part: duplication and divergence, our scenario 7.

Northern codfishes — de novo

The gadid AFGP gene shares no sequence identity with trypsinogen. It was traced to a minimal non-coding sequence, with the molecular mechanism of its conversion described — a documented de novo origin, and support for the proto-ORF route.

Two lineages, near-identical protein function, convergently evolved, one from an old gene and one from nothing. If you had only the proteins you could not tell them apart. Function does not carry origin information. Only the genomic neighbourhood does.

5 · Catching birth in the act

Zhao, Saelao, Jones and Begun (Science, 2014) sampled populations of Drosophila melanogaster and found 142 segregating and 106 fixed testis-expressed de novo genes, derived primarily from ancestral intergenic, unexpressed open reading frames — with evidence that selection contributed to their spread.

This is a different kind of evidence from anything comparative. A segregating de novo gene is one that some individuals carry and others do not: the birth is not a historical inference, it is a polymorphism you can genotype. If your system has population resequencing data, this is the strongest and most under-used line of evidence available.

6 · Vocabulary that is used inconsistently

TermWhat it should meanHow it gets misused
Orphan gene A gene with no detectable homolog in any other species Used interchangeably with "de novo gene". It is not — it is a search result, not an origin.
Taxonomically restricted gene (TRG) A gene detectable only within some clade — the honest, agnostic term Rarely misused, and it is the term to prefer when you have not established an origin.
Lineage-specific gene (LSG) Same as TRG, at whatever rank you specify Stated without the rank, so "lineage-specific" could mean species or class.
De novo gene Coding sequence derived from ancestrally non-coding DNA, demonstrated by synteny Claimed on absence of homology alone, which is the error the rest of this page is about.
Proto-gene A locus on the continuum: transcribed and translated, not established as a gene Used as a synonym for "gene I have not validated yet".
Novel gene Nothing specific. Avoid Everywhere, meaning any of the above.

The cheap discipline: call it a TRG until you have the syntenic evidence, then call it de novo. The upgrade costs you one alignment and buys you the whole claim.

Where these come from

ConceptSource
Homology detection failure; abSENSEWeisman, Murray & Eddy, PLOS Biology 18(11):e3000862, 2020 — "Many, but not all, lineage-specific genes can be explained by homology detection failure"
Synteny-based counterpointVakirlis, Carvunis & McLysaght, eLife 9:e53500, 2020 — "Synteny-based analyses indicate that sequence divergence is not the main source of orphan genes"
Proto-genes and the continuumCarvunis et al., Nature 487:370–374, 2012 — "Proto-genes and de novo gene birth"
Preadaptation; structural disorderWilson, Foy, Neme & Masel, Nature Ecology & Evolution 1:0146, 2017
Population genetics of de novo genesZhao, Saelao, Jones & Begun, Science 343:769–772, 2014
AFGP, de novo routeZhuang et al., PNAS 116:4400–4405, 2019 — non-sense to sense evolution in northern gadids
AFGP, duplication routeChen, DeVries & Cheng, PNAS 94:3811–3816 / 3817–3822, 1997 — notothenioid AFGP from trypsinogen
Phylostratigraphy itselfBarrera-Redondo et al., Genome Biology 24:97, 2023 — genEra

Volume and page numbers were taken from the publisher records at the time of writing; if you are citing these, pull the record yourself rather than trusting a web page — including this one.