data mining for gene function

From: Ramez Naam (mez@apexnano.com)
Date: Wed Apr 16 2003 - 00:52:09 MDT

  • Next message: Samantha Atkins: "Re: What Microsoft wants for your future"

    I have a question for the mathematically inclined on the list.
     
    In my book I'm trying to present a timeline for when we'll be able to
    /systematically/ discover the genes that impact various human traits.
    To do this I have a basic plot of # of human genomes likely to be
    sequenced by a given date, which in turn depends on the cost of
    sequencing a human genome at a certain time and some made up number of
    $$ spent on such a project. That much I have no problem with (though
    obviously it's just a toy system since I can't be fully confident of
    any of the numbers).
     
    The more challenging part is thinking about phenotypes for which there
    are many different genes involved, and where the genes may combine in
    a more-than-additive way.
     
    For example, consider the following entirely theoretical scenario:
     
    1) Humans who possess allele A1 of gene A have a normal distribution
    of IQ relative to the general population.
     
    2) Humans who possess allele B1 of gene B have a normal distribution
    of IQ relative to the general population.
     
    3) Humans who possess BOTH A1 and B1 have a statistically significant
    increase in their average IQ relative to the general population.
     
    So I want to be able to determine the number of humans whose genotype
    and phenotype we'd have to capture in order to find this effect. More
    generally, I'd like to be able to express that necessary sample size
    as a function of the size of the effect we're looking for and the
    number of different genes involved.
     
    I have a few ideas of where to start but suspect someone on here will
    have a better idea. Any thoughts?
     
    thanks,
    mez

    This archive was generated by hypermail 2.1.5 : Wed Apr 16 2003 - 01:02:15 MDT