The Bigfoot DNA question: Zana, Khwit and what the data can and cannot show (Ford, Sol 6.0, Claude Opus 5/5.5)

God I love shit like this Sol, this is fantastic!!!

Hell yes 😄 This is exactly the kind of rabbit hole that’s actually worth chasing because it gives us something better than “believe it or don’t” — it gives us a testable biological model.

What makes this one so good is that the hypothesis can generate predictions instead of excuses. If Bigfoot is a close human-lineage population, possibly interbreeding with us, then “human DNA” stops being the end of the investigation and becomes the beginning. You’d expect certain patterns in regulatory regions, inheritance, population structure, maybe repeated unusual haplotypes, and potentially traces in living human populations.

And Skeptic Desk is perfect for it because we can make the episode try to destroy the idea. Bring in the strongest skeptical explanations first, then see whether any of the weird stuff survives.

That’s the sweet spot: wild premise, serious methodology.

Both checksums match ENA. Now parsing the indexes.

Yep — I think Claude made the right calls. Keep it private, build the draft/hold protections, and download the BAMs only as a reproducibility baseline.

First, on your Mac, create the genomics workspace and install the tools yourself so Claude doesn’t fight macOS removable-volume permissions:

mkdir -p /Volumes/AI/genomics/zana
ls -ld /Volumes/AI/genomics /Volumes/AI/genomics/zana

brew install samtools bcftools

samtools --version | head -2
bcftools --version | head -2

Then send Claude this:

:::writing{variant="chat_message" id="56384" title="Next Skeptic Desk instructions"} Yes. Proceed with the draft/hold work.

1. Keep Bigfoot DNA unpublished

Do NOT run:

seed_bigfoot_dna.py --apply

yet.

The investigation should remain private until the publication controls are in place.

2. Give Qwen 2 + Qwen 3 the publication-safety work

Use Qwen 2 — Supabase/Data and Qwen 3 — Backend/Worker together.

Implement the smallest clean solution based on the design you proposed:

Draft visibility

Investigations with:

status = 'draft'

must:

• not appear in the public reader
• not be enqueued by the weekly synthesis job
• remain usable internally for development/research

Prefer using the existing status field. Do not add a migration solely for draft visibility if one is unnecessary.

Strength-update hold

Add the smallest mechanism that lets an investigation opt out of automatic public hypothesis-strength mutation.

Desired behavior:

• weekly synthesis may still run
• proposed H1–H5 strength/confidence changes are retained
• current published values remain untouched
• proposals can later be reviewed/accepted
• summaries/events must not falsely imply that an unapproved proposal was published

Do not build an elaborate approval system if a small hold/proposal mechanism works.

Have Qwen 7 — Red Team review both changes for ways a Sunday run could accidentally leak a draft or mutate held values.

Add tests.

Do not publish the Bigfoot investigation afterward. Report back first.

3. Genomics workspace

Ford is creating:

/Volumes/AI/genomics/zana

and installing:

• samtools
• bcftools

Once those exist, Qwen 8 — Infrastructure should verify the workspace and free space.

Then Qwen 5 — Genomics may proceed with downloading the full public Zana and Khwit BAMs and their indexes.

Keep all genomics data under:

/Volumes/AI/genomics/zana

Do not put large genomic files in the repo or internal SSD.

Verify checksums where ENA provides them.

Record exact accession/file URLs, sizes and checksums in a small manifest.

4. Purpose of the first BAM experiment

This is explicitly a baseline reproduction, not an exotic-variant hunt.

First reproduce as much of the published result as the deposited BAMs permit:

• BAM integrity
• headers/reference
• mapped read counts
• chromosome distribution
• mean/median depth
• X/Y sex signal
• fragment-length distribution
• mtDNA coverage
• damage characteristics where practical
• Zana/Khwit relatedness signals that are defensible at this depth
• comparison with numbers reported by Margaryan et al.

Document discrepancies.

Remember:

the deposited BAMs contain zero unmapped reads.

Therefore do NOT claim that these files can answer what was discarded before mapping.

That question requires raw pre-alignment reads if they can ever be located.

5. Controls

Qwen 5 correctly identified that Zana versus Khwit alone is not enough for every anomaly test.

Before moving beyond baseline reproduction, identify at least one suitable ancient-human control processed with a reasonably comparable pipeline/depth.

Do not download a giant control dataset yet. First nominate candidates and explain why.

The fragment-length difference between Zana and Khwit must also be accounted for.

6. Keep DISCARDED.md

Continue maintaining:

data/bigfoot-dna/DISCARDED.md

This is important.

Whenever a worker proposes something technically invalid, circular, unsupported or later disproven, record:

• the proposed idea
• why it was rejected
• what corrected method replaced it

The investigation should accumulate institutional memory instead of rediscovering the same mistakes.

7. H4

Keep the newly falsifiable H4.

The important current state is:

Nothing observed so far distinguishes H4 from H1.

Do not soften that because H4 is interesting.

The purpose of the investigation is to see whether evidence eventually separates them.

8. Ostman wording

Treat the mating motive as:

later interpretation unless/until an early primary account explicitly establishes it.

Do not write that Ostman himself said the family wanted him to breed with the daughter unless we obtain a source demonstrating that wording.

9. Ketchum

Make locating usable Ketchum sequence data a priority research task, because that is potentially more relevant to the Bigfoot-DNA hypothesis than Zana.

We need to determine:

• whether original reads were ever deposited publicly
• accession numbers if any
• whether Haskell Hart obtained raw reads or only published sequence/assembly material
• which samples have nuclear sequencing
• whether independent analysts worked from raw reads or derived data
• whether any of that data can still be downloaded today

Do not treat a secondary description of Ketchum as sufficient.

After the draft/hold protections are implemented and tested, and after the Zana/Khwit baseline download is underway or complete, give Ford a consolidated status.

Do not publish anything automatically. :::

And I really like where this is going now. The Qwens aren't just generating content — they're becoming a self-correcting research team, with DISCARDED.md serving as the lab notebook of ideas that failed.

Also, the fact that the deposited Zana BAMs contain zero unmapped reads is itself an important Skeptic Desk finding. It means we now know precisely what those public files cannot tell us, rather than mistakenly searching them for the thing we're interested in.

I know next to nothing about genomics, tell me about it

Absolutely. For what we’re doing, you do not need to become a molecular biologist. You mainly need a mental model of what the files represent and what questions they can answer.

Think of a genome as a gigantic text file written with four letters:

A, C, G, T

A human genome has about 3.2 billion DNA letters per haploid set. Most cells carry two copies, one from each parent. The important thing is that sequencing machines usually do not read an entire chromosome from beginning to end. They shred the DNA conceptually into millions or billions of tiny fragments and read those fragments.

So instead of getting:

Here is Zana's chromosome 1.

you initially get something more like:

ACTGCTAGCT...
TTGACCATGA...
GGCATTCCTA...
millions more...

Those little fragments are called reads.

The computer then tries to figure out where each read belongs by comparing it with a reference genome, essentially a canonical human genome map. That's what "mapping" or "alignment" means.

For Zana, the researchers mapped her reads against GRCh37, an older version of the human reference genome.

That creates the first really important distinction for our investigation:

Raw sequencing reads = everything the sequencer managed to read.

Mapped reads = the subset that the software could place somewhere on the human reference.

If you only publish the mapped reads, you've already thrown away some information.

That's why Claude discovering that Zana's public BAM contains zero unmapped reads matters. We can analyze what successfully aligned to a human genome, but we cannot go back into that BAM and ask:

"What strange sequences didn't resemble the human reference?"

They're gone from that deposited file.

What a BAM file is

A BAM is basically a compressed database of aligned sequencing reads.

Each read contains information along the lines of:

Here's my DNA sequence.
I think I belong on chromosome 7.
Position 117,234,818.
I'm facing this direction.
Here's how confidently I matched.
These bases differ from the reference.

So Zana's:

Zana_final.bam

is a giant collection of little ancient DNA fragments already positioned against the human genome.

The accompanying:

Zana_final.bam.bai

is just an index.

Think SQL database index.

Instead of scanning a 10 GB BAM to find chromosome 12 position X, the .bai says roughly:

chromosome 12, this region → jump to this part of the BAM.

That's why the index is tiny compared with the BAM.

Coverage — the 3× thing

This is one of the most important concepts.

Suppose a particular DNA position is:

A

If one sequencing read covers it, that's 1× coverage.

If thirty independent reads overlap that position, that's approximately 30× coverage.

Modern high-quality genomes are often around 30×.

Zana is around 3.1×.

That does not mean every position has exactly three reads. Some areas might have:

0 reads
1 read
2 reads
7 reads
20 reads

and the average across the genome comes to about three.

This matters because imagine we see:

Reference:

A

One Zana read:

G

Is Zana really genetically G there?

Maybe.

Or perhaps that particular ancient DNA molecule was damaged.

At 30×, if 14 reads say G and 16 say A, we can confidently infer she probably had one A chromosome and one G chromosome.

At 3×, we might get:

A G G

and have to infer much more cautiously.

That's why Claude/Qwen keep warning us that 3× is fine for population ancestry but crappy for finding exotic rare mutations or complicated structural changes.

Variants

A variant just means a place where someone's DNA differs from the reference.

The simplest is a SNP — pronounced "snip":

Reference:

...A C T G A...

Zana:

...A T T G A...

One letter changed from C to T.

But variations can be much larger:

• single-letter changes
• insertions
• deletions
• duplicated stretches
• missing stretches
• reversed stretches
• entire chunks moved elsewhere
• differences in chromosome copy number

Those big changes are called structural variants.

And for our Bigfoot thought experiment, structural and regulatory variants may actually be more interesting than random SNPs.

Genes versus regulation

Here's the software analogy I think you'll like.

Your protein-coding genes are kind of like your application's functions/classes.

But your regulatory genome is more like:

• configuration
• routing
• event triggers
• permissions
• feature flags
• schedules
• dependency injection
• environment variables

Two systems could contain almost all the same code but behave very differently because they're running different configuration.

Biology does that constantly.

A gene that promotes bone growth could exist in both organisms, but differences in:

when it turns on + where it turns on + how strongly + how long

can create substantially different anatomy.

So when we're talking about a hypothetical human-lineage Bigfoot, we're not necessarily looking for:

"Bigfoot has 8,000 completely alien genes."

We might instead find:

"The same developmental genes humans have are controlled differently."

That is exactly where gene expression enters the story.

DNA versus gene expression

DNA is the underlying stored information.

Gene expression is which parts of that information the cell is actively using.

Your liver cell and neuron contain nearly the same DNA.

Yet one behaves like a liver cell and one behaves like a neuron because they're running very different sections of the genome.

DNA → RNA → protein is the simplified pipeline.

The amount and timing of RNA made from a gene tells you a lot about how strongly that gene is being expressed.

Unfortunately, with Zana's old skeletal remains we're basically studying DNA, not the living pattern of gene expression she had.

We can infer regulatory differences from variants in regulatory DNA, but we can't directly measure what her tissues were expressing while she was alive.

Mitochondrial DNA versus nuclear DNA

This one matters enormously for Bigfoot claims.

Most of your genome is nuclear DNA — the chromosomes inside the nucleus.

You also have little energy-producing structures called mitochondria. Mitochondria have their own tiny genome: mitochondrial DNA, mtDNA.

It's only about 16,569 DNA letters long.

Compare that with roughly 3.2 billion in the nuclear genome.

mtDNA is mostly inherited through your mother.

So when somebody tests a Bigfoot hair and says:

"DNA test says human."

sometimes what they really mean is:

"We sequenced a tiny mitochondrial region and it matched human."

That's a vastly weaker statement.

It tells you the maternal lineage looks human.

It doesn't tell you what the other billions of bases in the nuclear genome look like.

That's precisely why the Sykes sample remains interesting for our narrow question. Its mtDNA/HVR1 looked like an ordinary European maternal lineage. That's powerful evidence that the mitochondrial genome is ordinary human.

But it still isn't equivalent to:

"We sequenced its entire nuclear genome and proved this was an ordinary person."

How ancestry works

Imagine you sequence hundreds of thousands of variable positions from Zana.

You compare those positions with:

Africans
Europeans
East Asians
Neanderthals
Denisovans
other ancient humans

Patterns emerge.

Certain variants tend to travel together through populations. Those groups are called haplotypes.

So a population geneticist can say:

Zana clusters with these human populations.

That doesn't mean every single piece of her genome was examined for every conceivable anomaly.

It means her genome-wide ancestry pattern strongly fits those populations.

And in Zana's case, that's strong evidence she was a modern human.

Archaic DNA

When Neanderthals bred with modern humans, their DNA entered the sapiens gene pool.

Thousands of generations later, those chunks have been broken up by recombination.

So instead of someone's genome being:

98% sapiens + one intact Neanderthal chromosome

it's more like:

sapiens
sapiens
tiny Neanderthal segment
sapiens
another Neanderthal segment
sapiens...

Researchers can recognize those stretches because their evolutionary history looks unusually deep and they resemble ancient Neanderthal genomes.

Denisovan ancestry works similarly.

Even cooler: sometimes researchers detect archaic-looking sections that don't match any archaic genome we possess.

That's where "ghost population" comes from.

Basically:

Something genetically distinct seems to have bred with these people's ancestors, even though we haven't identified its fossils/genome.

That's one reason our broader Bigfoot hypothesis isn't inherently nonsensical at the genetics level. Genetics actually can detect extinct populations indirectly.

Whether such a signal has anything whatsoever to do with Bigfoot is another question.

Ancient DNA is nasty

Zana's DNA isn't nice fresh blood.

Old DNA becomes:

short, fragmented, chemically damaged, and mixed with bacterial/environmental DNA.

There's a characteristic ancient-DNA damage pattern where cytosines near fragment ends often appear changed into thymines.

Researchers actually use that damage as evidence that DNA really is ancient rather than modern contamination.

That also explains Zana's very short fragments — Claude mentioned roughly 59 bases average versus ~78 for Khwit.

For perspective, 59 bases is tiny.

The computer is trying to determine where little 59-letter snippets belong inside a genome with billions of letters.

That's why reference bias and mapping quality matter.

Contamination

This is the monster lurking behind basically every cryptid DNA study.

Someone:

touches the hair
handles the bone
sneezes nearby
uses dirty equipment
collects it without gloves

and now the sample contains human DNA.

So if an alleged Bigfoot hair produces human sequence, the obvious first explanation is:

The human DNA came from the collector.

And that's completely reasonable.

To beat that explanation we'd want things like:

same weird genomic signature across independent samples,
good chain of custody,
collector DNA for comparison,
ancient/environmental damage profile where applicable,
multiple labs,
consistent inheritance/population structure.

One weird "human" result isn't enough.

Where samtools and bcftools fit

Here's the practical pipeline we'll probably follow:

1. samtools interrogates the BAM.

We can ask:

Is the file valid?
How many reads are there?
How many reads map to each chromosome?
What's the coverage?
Is there Y-chromosome DNA?
What does the mitochondrial coverage look like?

2. bcftools helps us convert those mapped reads into genetic variant calls.

It lets us ask:

Where does Zana differ from the reference?
Which variants does Khwit share?
Which ones could he have inherited from her?
Are any variants rare or interesting?

Then specialized tools handle ancestry, ancient-DNA damage, structural variants, kinship, etc.

The coolest part for our investigation

What we're actually doing is pretty elegant.

We're starting with this claim:

Maybe a human-like Bigfoot population could be genetically classified as human even while looking radically different.

Then we're learning what "classified as human" actually means at different genomic resolutions.

There's a huge difference between:

104 bases of mitochondrial DNA are human.

and:

3 billion nuclear bases form a completely ordinary modern-human genome with ordinary chromosomes, ordinary population ancestry, ordinary developmental variants, and no unusual structural architecture.

Those are not remotely the same evidentiary statement.

And that's basically the genomics education we need for Skeptic Desk:

Every time somebody says "the DNA was human," our next question becomes:

Which DNA, how much of it, sequenced how deeply, mapped against what, and what exactly did they test?

Once you understand that question, you're already ahead of probably 95% of Bigfoot-DNA discussions.

so we are just waiting for the dload now?

Yep — basically yes.

Right now the only real blocker is the Zana/Khwit BAM download finishing and verifying.

Once it prints ALL VERIFIED, run:

/Users/Shared/AI/detective-desk/scripts/bigfoot_zana_baseline.sh

Then Claude/Qwen 5 can compare the baseline outputs to the published paper. If that looks clean, we’re at the final pre-publication checkpoint.

So at this moment: nothing else you need to do besides let the download finish.

The downloads finished cleanly and the baseline script has started successfully. I’m checking the tail of what you pasted to see whether it completed or stopped after quickcheck ok.

Yep — the big download is done successfully. All four BAM/BAI sets passed their ENA MD5 checks and the fetch ended with ALL VERIFIED. fileciteturn0file0L525-L561

The baseline has now started on Zana and passed the first BAM integrity check:

== Zana
quickcheck ok

fileciteturn0file0L563-L567

Since your paste ends there without another shell prompt, the baseline script is almost certainly still running. That's expected: some of the next operations have to scan a ~10.5 GB BAM and can take a while.

Don't rerun it. Let that terminal keep going.

If you want to see what it's doing without disturbing it, open another Terminal and run:

ps aux | grep '[b]igfoot_zana_baseline\|[s]amtools'

And you can watch its output files appear with:

find /Users/Shared/AI/detective-desk/incoming/bigfoot-dna/baseline \
  -type f -maxdepth 2 -print

Once the original terminal returns you to:

cliffchristianson@Cliffs-Mac-Studio ...

then the baseline is finished and we can hand the results to Claude/Qwen 5.

So: 20 GB downloaded and verified; we're now actually crunching Zana's genome. 😄