Hacker News Reader

Comment by causality0 | Hacker News Reader

I can't find any scientists claiming that sequencing 700,000 year old DNA means the half-life study was incorrect, or even that it was incorrect for any other reason. Do you have sources?

View on HN

It's math. If the half-life was 500 years, there would be nothing left to sequence after 700,000 years.

Kilograms of biological material that's been frozen for 0.7MY will contain many orders of magnitude more intact DNA than a few grams of bone marrow that's been in assorted temperatures for a hundred times as long.

700,000 / 500 = 1400 half lives, or 2^(-1400).

I'll let you do the math on how much would remain if we had 1 million tons of horse DNA with a 500 year half life. I'll even let you assume we have 1 trillion tons of horse DNA.

The half-life paper itself cites a readability horizon of 1.5 million years for even non-permafrost DNA. Regardless of what the degradation rate is, reading the genome of a whole frozen animal from 0.7MY ago is much easier than reading the genome from a tiny piece of unfossilized gel that's a hundred times older.

I know that feeling, my man. My shelves are covered in Jurassic Park memorabilia too. It's just not going to happen, though.

But then what's the point of the study? DNA has a half life of 500 years except when it doesn't? If they say 500 years and then you can retrieve DNA from samples 1000 times as old, who's to say they can't go further by another factor of 100?

I never said it was impossible. I meant that based on the fact DNA destabilizes over time even without an external destructive force, that is, it has a half-life, it's supremely unlikely useful genomes will ever be derived from 50+ million year old fossils.

DNA obviously does not behave like radioactive isotopes, so the half-life is dependent on a lot of external factors. The study presents a model for estimating the half-life of DNA for a given storage temperature assuming otherwise similar conditions to their reference samples. That's obviously useful for anyone working with ancient DNA, since it helps guide attention to samples that could potentially contain recoverable DNA.

The 521 years is just the half-life they estimated for their samples. Those are the empirical results, so it makes sense to highlight those in the title, as opposed to estimates from their model.

I don't mean to be rude, but please read the fine manuscript. It's open access, so you don't need to pay

If you have a bunch of observations, you can fit an exponential through them. You can do that in Excel. Here, let me just quickly do that for the half life of the buildings in Europe. ... Ok, done, it's 63 years. Should I publish an article? What conclusions does one draw? The usefulness of a model is how well it can extrapolate. 63 years means I should not find any buildings that are 2000 years old, and still the Pantheon is standing there just fine. Then what exactly is the claim? That it's possible to fit an exponential, but the exponential can't be used for inferring anything outside of my training sample?

And, no you are not being rude. But since you seem to have read the article, can you give me a few pointers why you think it's a good investment of someone else's time to go ahead and read it too? And don't be shy to use Machine Learning or Statistics jargon, we are on Hacker News here, everyone understands concepts like in-sample, out-of-sample, goodness of fit, overfitting and such.

You disagree with the 521 paper, but I'm not sure what your position is. Are you positing that DNA will not destabilize on its own unless subject to a particularly hostile environment? Are you saying it has a half-life but it's so much larger than 500 years that you expect usable DNA to be found in unfossilized dinosaur bone marrow?

Reply on news.ycombinator.com