During the 2000 presidential election I was living in a fraternity house with a roommate serving in the Air Force.  When Bush was projected to be the winner, he jumped up and yelled,

"We're gonna have toys!  We're gonna have more toys!"

As he predicted, the next 8 years resulted in plenty of work and funding for the flyboys.  With the election of Barack Obama, perhaps us geneticists can start doing the same kind of dance.

 Sure, the election of a Democrat doesn't guarantee we'll get the funding we want.  For that to happen, we'd have to be a larger group of folks and have a less predictable voting record.   But, one legacy of the Bush administration is inadequate funding to the life sciences.  It's a fair wager that it won't get worse under Obama.  If his promises and track record are good indicators, funding should get a lot better, as indicated in this piece on msnbc:  expect funding to go up for research relating to "personalized medicine."

How well can we expect "personalized medicine" to succeed?

Personalized medicine entails using the genetic variation present in individuals to predict their responses to treatments.  For example, a certain variant of a gene might predispose some individuals to a severe side effect in response to a drug.  For personalized medicine to work, we must be able to identify specific genetic changes responsible for clinical variation, AND predict their effects. 

This prediction, in a word, is difficult.  Because statistics and large sample sizes must be used, the effects of genetic variation is studied in populations and usually measures only the average effect of a variant across the population.  Personalized medicine, as the name suggests, focuses instead on individuals;  knowing the effect of a genetic variant "on average" is not acceptable.  There is a fundamental challenge in advancing  from the study of the average effects of a genetic variant to predicting specific effects in individuals.  Based on volumes of data in organisms from yeast to humans, I believe that two of the most pressing issues here are the prevalence of genetic interactions and the effects of rare genetic variants.

A genetic interaction means that the effects of a variant depends on the state of other variants in the genome.  For example, a specific variant (also called an allele) of  gene X might cause a severe side effect reaction to a drug, but this reaction might be dependent on presence of a second interacting variant in gene Y.  If that interacting Y allele is unknown, then we will observe that some individuals carrying the "bad" allele of gene X have a severe reaction, but others are fine.  To truly  predict the effect of variation at gene X, we have to know and understand gene Y as well.  The response of individuals to the drug requires knowing the genotype at two genes, not just one.  In practice, genetic interactions can occur between several genes at once.  If interactions are common--and they appear to be--then to understand the effect of a variant, we must also know all of its interacting partners.  This makes the problem exponentially more difficult.

Another complication is that the frequencies of alleles can fluctuate.  The "bad" allele of gene X might be very common, but the "bad" allele of gene Y could be very rare.  In this case the interaction would undetectable in a population even though it has a very real and painful effect in the few individuals that unlucky enough to carry the unfavorable combination.  (I wrote here about a genetic interaction identified that effects male pattern baldness).

In the limit case, interactions could occur due to new mutations present only in a single individual.  If so then personalized medicine might be hopeless:  since the new allele has never been observed before, there is no way to predict its effect.  There is mounting evidence that new mutations are responsible for a significant fraction of cases of schizophrenia.

Looking Forward

To understand how well personalized medicine might work, we must understand how much of the genetic variation in nature is due to interactions and rare variation.  We simply don't have answers to those questions yet.  I argue that research should focus on two avenues to get at this fundamental challenge.  First, we need to develop analytical tools specifically designed to detect genetic interactions in humans in a systematic manner.  At present those types of tools are not commonly used in studies of genetic variation because there just isn't enough statistical power using current designs. 

Second, we need to answer the fundamental questions in model organisms.  Quantitative genetic theory predicts that both the prevalence of genetic interactions and the effects of rare variants will depend largely on the type and strength of selection present in a population.  By measuring the effects and interactions of genetic variants in more tractable systems like yeast, roundworms, flies, and mice, we can test the predictions of theory and carry what we learn into the design of studies in humans.