Mysterious mutants: 13 masked people should have devastating diseases but don’t | Hacker News Reader

Mysterious mutants: 13 masked people should have devastating diseases but don’t

arstechnica.com

40 pointswalterbell12 comments

While interesting the 13 people could have a mosaicism or a mistake could have been made. There is no way to tell as the study was anonymised. So until another study is done which offers traceability back to the 'mutants' we'll never know if there is anything useful here or not.

We should use this information to genetically engineer disease resistant people.

I say "we should" because if we don't the <insert technically advanced nation of your choice here> will.

Would it be possible to use this data to modify the DNA of a sperm / egg cell combo? Are we there yet?

I read somewhere that some team already was able to create viable embryos with modified genetic materials albeit the embryos were terminated after reaching a certain stage of replication.

most recently http://www.nature.com/news/second-chinese-team-reports-gene-...

added http://snpedia.com/index.php/Rs333 for HIV immunity.

the embryos were pre-non-viable.

One of the most difficult parts is both the desire (and possibly need) to gather more DNA from a pool as large as possible, with the option for re-contact for further investigation while also keeping that data safe and anonymous to prevent this being an incredible valuable target for both governments and corporations.

I think I watched a Bayes Probability vid on Youtube that basically says that 13 out of the 600,000 is most probably false results..

SciHub link to research paper: http://sci-hub.io/http://www.nature.com/nbt/journal/vaop/ncu...

Fear not as we'll eventually track down these 13 people once we combine these disparate DNA data sources. They might not have identity information in this particular dataset, but if one finds that same sequence in another dataset that does have that information, it's not too hard to piece these together.

Best part is that we don't even need it to be exact! We can look for their relatives instead and use that to narrow the search: http://fusion.net/story/215204/law-enforcement-agencies-are-...

Even if the sequencing process has very high accuracy, the conditional probability these people having the genes but not disorders could be low.

By Bayes Rule:

PR(Gene & no disorder | disorder result) =

Pr(disorder result | gene & no disorder) * Pr(gene & no disorder)

/ Pr(disorder result | gene & no disorder) * Pr(gene & no disorder) + Pr(disorder result | no gene & no disorder) * Pr(no gene & no disorder)

Even if the test is incredibly accurate and so Pr(disorder result) is much higher for people with the gene, this may be outweighed by the relative likelihood of having the disorder given you have the gene (this could plausibily be 1!), making the conditional probability very low.

But will the test always give a false positive for some people or is it usually more random?

If it is, then just test the sequences again.

Yeah, if you have tests without strong correlations then your approach is a good one. My point is that when your test returns a result which had low prior probability (even if the test is accurate) you should often retain a lot of your scepticism. In this context, it may well be that many of the people identified didn't have the genetic disorders indicated...

23andme keeps telling me that I am most definitely having lactose intolerance.

But I don't.

Reply on news.ycombinator.com