The 1000 Genomes Project

jd writes "An international consortium of specialists in genetics has announced the 1000 Genomes Project, in which at least 1,000 people from around the world will have their genomes fully sequenced as part of an effort to discover the relationship between genetics and disease. At present, over 100 regions of DNA are known to be related to illnesses, but the maps that exist are vague and are drawn from an extremely small population pool. According to the article, this results in the need for slow, expensive, and laborious studies to pinpoint causes, especially for rarer conditions. This project aims to find conditions that might only appear once in every 2,000 people (though how they intend to do that with half that number is unclear). The researchers hope to massively speed up the diagnosis of genetically linked illnesses and to improve the reliability of such diagnoses."

Comments

Selection

[mastershake_phd]

This project aims to find conditions that might only appear once in every 2,000 people (though how they intend to do that with half that number is unclear).
 
Well, they could sequence the DNA of people known to have rare diseases.

1 in 2000 people

[rsidd]

This project aims to find conditions that might only appear once in every 2,000 people (though how they intend to do that with half that number is unclear)

Let's try to make it clearer, then.

The probability that a given condition appears in an individual is 1 in 2000, or 0.0005. The probability that it does not appear in that individual is 0.9995. The probability that it does not appear in any of 1000 individuals is 0.9995^1000 = 0.6 approximately; and the probability that at least one of the 1000 individuals has it is 0.4. Not bad at all. (If you used 2000 people, the probability that at least one of them would have it would improve to about 0.6.)

Suppose you aren't interested in just one conditions, but in lots of conditions -- say, ten of them. The probability that at least one individual would have at least one of those conditions is 1 - 0.9995^(1000*10) = 0.993 == ie, practically certain.

They really ought to teach basic probability theory in schools...

Re:1 in 2000 people

[nacturation]

This project aims to find conditions that might only appear once in every 2,000 people (though how they intend to do that with half that number is unclear)

Let's try to make it clearer, then.

The probability that a given condition appears in an individual is 1 in 2000, or 0.0005. The probability that it does not appear in that individual is 0.9995. The probability that it does not appear in any of 1000 individuals is 0.9995^1000 = 0.6 approximately; and the probability that at least one of the 1000 individuals has it is 0.4. Not bad at all. (If you used 2000 people, the probability that at least one of them would have it would improve to about 0.6.)

Suppose you aren't interested in just one conditions, but in lots of conditions -- say, ten of them. The probability that at least one individual would have at least one of those conditions is 1 - 0.9995^(1000*10) = 0.993 == ie, practically certain.

They really ought to teach basic probability theory in schools...

Your post is like that scene in Indiana Jones with the guy making some really impressive sword moves, looking all menacing... while Indy just pulls out his revolver and shoots him. You could get a whole room full of geeks cranking numbers and arguing over how many people they would need to find in order to exceed a particular probability that any one participant has Lou Gehrig's disease, while a simpler person would leave the room, come back the next day, and say "Hey guys, meet my neighbor Bob... he has Lou Gehrig's."