Hacker News Reader

Comment by ginko | Hacker News Reader

DNA has a ~500 year half life.

For a 300,000 year old copy of a genome to survive there’d need to be 2^600 copies to begin with.

But thats of the whole DNA molecule. When it breaks down, the small chunks leftover still contain fragments of information.

This is not impossible. 2^600 is approximately 1e19. Considering there are 92 strands of DNA in a single cell, and around 37 trillion cells in a human body, a quick calculation means you need to dig up around 3000 bodies. Of course, this isn't exact (I haven't taken soft tissue decay into account, for instance), but the job is made somewhat easier by being able to piece together broken strands of DNA

> 2^600 is approximately 1e19.

No, it's approximately 1e180. You're off by 160 orders of magnitude.

But those 3000 bodies don't have exactly the same genomes obviously. In fact half of them won't even share one of the chromosomes.

It'll be hard to recombine this into a clear picture as you won't know which sequences are common and which are individual.

That isn’t a problem at all, you will get a consensus sequence along with additional information about which variants differ in the population. That is even better info than a single perfect genome sequence.

Not impossible but there are risks in that plan. We could be in a similar problem than chatGPT is having, where part of the combinations would end being wrong and generating newer info, that would reinforce the error.

Reply on news.ycombinator.com