A two-axis functional and mitochondrial-origin annotation of 189 ancient human NUMTs, and an ancestral-reconstruction test showing SLAIN1's lost cross-species alignment eroded rather than vanished.
Jayden · Frith Lab, University of Tokyo · UTSIP 2026
Every one of your cells keeps two completely separate sets of DNA, in two different places.
Once in a while a scrap of mitochondrial DNA gets copied and pasted into the nuclear DNA. That stray copy is a NUMT: a genomic fossil, because many were pasted into our ancestors long ago and have been carried in the genome ever since.
Right after the copy-paste, the nuclear copy is letter-for-letter identical to the mitochondrial original.
We line up two stretches of DNA and see how well they match.
Two ways to find them: direct alignment (mitochondria vs nucleus) and cross-species alignment (one animal's nucleus vs another's). Together they turned up all 189.
Prof. Frith's lab already found 189 of these ancient NUMTs in our DNA. The big question: are they just random junk, or do any still matter?
We'll answer each one as we go. (Numbers and details are in my back pocket for questions.)
Objective 1
I built a pipeline that finds each NUMT's origin point in the mitochondria, then labels which gene it came from, matched against the reference annotation of the mitochondrial genome.
Spin the wheel: every spin lands on a part of the mitochondria. Keep spinning, your tally (top) fills in to match what we actually found (bottom).
Objective 1 · why
The wheel is really the mitochondria: each spin is a random copy-paste that can end up in the nuclear DNA as a NUMT.
So the NUMTs we find just mirror this random, size-driven process, which is why our real distribution matched the spins.
Objective 1 · a check
I ran the same pipeline both ways and compared the results: does how we found a NUMT change the picture?
direct alignment (in us) cross-species alignment (via other animals)
Our results show direct alignment and cross-species alignment give the same spectrum, so cross-species alignment isn't skewed. We can trust it to catch the faded NUMTs that direct alignment misses.
Objective 1 · the answer
Where did each come from: random, or chosen?
Bigger genes are just bigger targets, so which parts become NUMTs is pure chance. And both ways of finding them agree, so it's real, not an artefact of how we looked.
Objective 2
The pipeline also records where each NUMT sits in the nuclear DNA and classifies what's there (a gene, a control switch, or quiet DNA) from the genome's annotation.
found directly (mito vs nucleus) found cross-species (through other animals)
About half land in DNA that does something: regulatory, active, even exons, not just quiet space. Both ways of finding them give the same split, so neither method is biased toward one kind of region: the pattern is real.
Covered at the mid-presentation and in your handout, so I'll keep this brief and focus on Objective 3.
Objective 2 · the answer
Objective 2 asked where all 189 landed, and whether it matters. Here's the catch: most of our genome is junk DNA, the long stretches between and inside genes that don't do a job.
So by the same random logic as the spin-wheel, most NUMTs should land in that junk, shouldn't they?
Objective 2 · the answer
In DNA, the parts that matter change slowly; the parts that don't change fast, because natural selection keeps what's important. Think of a country's laws versus a time you agreed to meet a friend: the important one is harder, and rarer, to change.
Objective 3 · the hero
Meet SLAIN1: of all 189 NUMTs, one of the oldest (traceable back ~160 My) and one of the slowest-changing.
The mouse carries the same NUMT, but it has mutated so much it no longer lines up with other animals' copies. Did they ever match: were they the same to begin with? That's what we set out to test.
SLAIN1’s mouse copy, lined up against our oldest cousins: the opossum and platypus, ~160 My away. (Nuclear DNA vs nuclear DNA.)
Today’s mouse copy, lined up against the cousin.
Reconstruction by genancestor, Prof. Frith’s ancestral-DNA tool.
Objective 3 · the answer
Can we bring a faded NUMT back?
The match the fast-evolving mouse had lost came straight back the moment we reconstructed its ancestor. And SLAIN1 sits in your genome too: faded, but still there.
What comes next
We just watched reconstruction bring back a NUMT the mouse had lost. So here's the next question: if we rebuild the ancestral mitochondrial genome and run direct alignment with it against the nuclear DNA (the same way we first found NUMTs), could it uncover ones today's mitochondrial genome can no longer find?
What comes next · the pilot
We aligned with the reconstructed ancestral genome and compared what it finds against Prof. Huang's known NUMTs.
| animal | found by both | only the ancestral search found | only today's search found |
|---|---|---|---|
| human | 978 (94%) | ~74 ← new leads | 80 |
| mouse | 229 | 39 | 64 |
| sloth | 295 | 42 | 50 |
In us, it re-found 94% of the known NUMTs (so it works) and flagged ~74 spots today's genome misses.
Along the way: a labelled catalogue of all 189 ancient NUMTs, and a way to tell true loss from mere fading.
A 160-million-year-old genomic fossil, brought back into focus.
Audience questions appear here live.