Why This Site Has Almost Nothing on Huntington's, ALS or Cystic Fibrosis

People arrive at a genetics site with the hardest questions first. Huntington’s disease. ALS. Duchenne. Cystic fibrosis. The diseases where somebody already knows the gene, and still nothing can be done.

This site has almost nothing on any of them, and that is not an oversight. It is worth explaining properly, because the reason turns out to be more interesting than the gap.

First, the actual number

Of the 660 variants published here today:

• Huntington’s disease — 0
• ALS and motor neuron disease — 0
• Parkinson’s disease — 0
• Duchenne muscular dystrophy — 0
• Spinal muscular atrophy, fragile X, Tay-Sachs, prion disease — 0
• Cystic fibrosis — 1, and it is a modifier of how severe the disease is, not the gene that causes it
• Alzheimer’s disease — 2

Those zeros are not all the same kind of zero, and telling them apart is the point of the rest of this page.

Why a catalogue like this cannot see most of them

Everything here comes from genome-wide association studies: research that takes hundreds of thousands of common single-letter differences, compares them between people with a condition and people without, and reports the ones that differ more than chance allows.

That method has one requirement that is easy to miss. The thing causing the disease has to be common, and it has to be a single letter. Most of the diseases above are neither — and the word most is doing real work there, which we come back to in a moment.

Huntington’s is caused by a stretch of DNA that repeats too many times. ALS’s commonest inherited cause is a different stretch that repeats too many times. Duchenne is usually a chunk of a gene missing altogether. Cystic fibrosis is caused by mutations that are individually rare, in one gene, in people who mostly have two broken copies.

A scan of common single-letter differences will not find any of that — not because it was done badly, but because it is not what the instrument measures. Ours is a catalogue of what that instrument found. Running it for another ten years would not add a single Huntington’s variant.

So the honest answer to “why isn’t Huntington’s here” is: because this is the wrong kind of catalogue for it, and pretending otherwise would be worse than the gap.

Except Parkinson’s, which is our backlog and not the method’s limit

Parkinson’s disease is in the list above and does not belong in the argument just made. It is a common-variant disease, and it has one of the largest genetic studies in neurology behind it.

A 2019 meta-analysis looked at 7.8 million single-letter positions in 37,688 patients, 18,618 people whose parent or sibling had the disease, and 1.4 million controls. It found 90 independent genome-wide significant signals across 78 regions — and those 90 together account for 16 to 36% of the inherited risk of Parkinson’s, depending on how common you assume the disease to be.

Every one of those is exactly the kind of finding this catalogue exists to hold. We have none of them, and the reason is unglamorous: our collector works through the public GWAS Catalog a page at a time and has reached page 45 of 23,841. That is under 0.2% of it. It has not got to Parkinson’s yet.

That is a backlog, and backlogs fill. Filing it next to Huntington’s would mean calling a gap in our collecting a limit of the method — which is precisely the sloppiness this page is complaining about, so it gets its own heading instead.

The wrong conclusion to draw

It would be easy to read all that as “genetics has little to offer these diseases”. That is the opposite of true, and Huntington’s is the proof.

Huntington’s: thirty-three years of knowing exactly what is wrong

In 1993 a consortium found the gene. They looked at 75 families and found the same thing in every one: a CAG sequence, repeated too many times, in a gene nobody had seen before.

It is about as clear as human genetics gets. Enough repeats and the disease will come; the more repeats, the earlier. No environment saves you, no lifestyle changes it, and a parent with it passes it to each child with probability one half.

Thirty-three years later there is still no treatment that stops it.

That is the fact that makes people ask what genetics is actually for. Here is the answer.

You cannot ask who. So they asked when.

A genome-wide study needs two groups to compare: people who get the disease and people who do not. For Huntington’s that comparison is impossible. Everyone with a long enough repeat gets it. There is no resistant group to find.

So the field changed the question. Not who develops Huntington’s, which the repeat already answers, but when does it start — because that varies by years between people whose repeats are the same length, and nobody knew why.

That question turned out to be answerable.

A scan across the genomes of thousands of patients found a locus on chromosome 15 with two independent effects, one delaying onset by 6.1 years and one bringing it forward by 1.4, and a locus on chromosome 8 bringing it forward by 1.6.

Six years is not a cure. It is also not nothing: it is the difference between a diagnosis at 44 and a diagnosis at 50, and the people who have those six years already have them, inherited, without anyone doing anything.

The authors put it in a sentence worth repeating: disease modification in humans occurs in nature. Somewhere out there the thing a drug is trying to do is already being done by a variant. Find it, and you have a target that has been validated in humans before anybody spent a penny.

And the target was not the one everybody expected

For thirty years the assumption was that the repeat matters because of what it makes: a long run of CAG produces a long run of glutamine in the protein, and the long glutamine stretch does the damage. Every therapeutic idea followed from that.

The onset study pointed somewhere else entirely. The genes it landed on — including MLH1 — were DNA repair genes. Not protein disposal. Not neurotransmitters. The machinery that maintains DNA.

The follow-up made the case directly. The timing of onset is set by the length of the uninterrupted CAG repeat, not by the length of the glutamine stretch it encodes. The two can be separated, because the repeat sometimes contains a CAA interruption: a different triplet that codes for the same glutamine. Same protein, different DNA — and it is the DNA that tracks with when the disease starts.

The clearest evidence is at the bottom of the figure above. In people who have lost that CAA interruption, onset comes on average 20.4 years earlier than their measured CAG length predicts. The protein is unchanged. The DNA is less stable.

What that means, in one paragraph

The repeat is not a fixed inheritance that sits there doing damage. It grows — inside cells, over a lifetime, in the neurons that are eventually lost. A person inherits a repeat of one length and carries a longer one in the affected part of the brain decades later. The DNA repair genes are modifiers because they set how fast that happens.

Which turns the therapeutic question inside out. The goal is no longer only to clear a damaged protein. It is to stop the repeat expanding — and unlike the protein, that process has known, targetable machinery.

There is a detail here that should be uncomfortable for anyone who thinks progress is fast. The 1993 paper had the word unstable in its title. It compared the repeat to fragile X and myotonic dystrophy, both of which were already known to be unstable. The clue was in the first sentence of the discovery, and it took twenty-six years for instability to move from a footnote about the sequence to the mechanism of the disease.

ALS, briefly, for the same reason

The commonest inherited cause of ALS is also a repeat — six letters, GGGGCC, in a gene called C9orf72. It was found in 2011, in 23.5% of families with inherited ALS and 11.7% of families with frontotemporal dementia, which is part of why those two diseases turned out to be related at all.

Same shape of story: the single most important genetic fact about the disease is invisible to the method that built this site.

And cystic fibrosis, which is the reason not to be fatalistic

Cystic fibrosis was, in living memory, exactly the kind of disease this article is about: gene known since 1989, cause understood, nothing to be done but treat the consequences.

It is not that any more. There are now drugs that go to the faulty protein itself and make it work better — and which drug helps you depends on which mutation you have. Roughly 90% of people with CF carry at least one copy of the same mutation, Phe508del, and the combination therapy tested for that group in 2019 is the reason the outlook for a child diagnosed today is not the outlook from twenty years ago.

That is what genotype-directed treatment looks like when it arrives: not a risk score, but a prescription that is chosen by which letters you have.

About testing, said plainly

If someone has read this far because Huntington’s is in their family, this part matters more than the rest of the page.

Predictive testing for Huntington’s in someone without symptoms is done through a counselling protocol, over more than one appointment, and that is not bureaucracy. The result cannot be undone, it says something about siblings and children who did not ask, and it is information for which there is currently no treatment. People who go through the process sometimes decide not to have the test, and that is a legitimate outcome of it rather than a failure.

Nothing on this page is a route around any of that. The way to that decision is a clinical genetics service, and it is worth saying that they are good at this.

One concrete technical point, because it settles the question for our own readers. Consumer genotyping data cannot answer this. An array reads single letters at fixed positions; it does not measure how many times a sequence repeats. The same is true of the raw data files our own desktop reader opens. Whatever anyone tells you they found in their raw data, the CAG repeat is not in there — it takes a different test, ordered deliberately.

What we will do about the gap

We are not going to fake coverage by attaching thin common-variant findings to serious diseases, and we are not going to leave the questions unanswered either.

Where there is something honest and useful to write — a mechanism, a modifier, a piece of evidence that changes what a family is told — we will write it, and it will look like this page rather than like a risk lookup. Where the truthful answer is “this is not knowable from a genotype”, that is what the page will say.

The site’s promise has never been that it holds everything. It is that what it does hold is true, and that where it is empty, it says why.