The problem is this: it's becoming cheaper for any of us to go out and get ourselves genotyped. Companies like 23andMe, deCODEme, and Navigenics will give you a read out of your DNA at hundreds of thousands of positions in your genome. Ultimately the goal is to be able to take those DNA results to your doctor and say, what does this all mean?
Right now we're limited in what we can say. It is hard to tell whether a DNA variant at a certain position is a harmful one, or just harmless variation. Kumar tackled this problem by looking for traits that distinguish known mutations that are associated with a disease from general DNA variation.
In particular, Kumar looked at evolutionary traits. By comparing human DNA with the comparable genome sites in the mouse, chicken, and puffer fish genomes, he found that human disease mutations tend to occur at sites that have changed very little during the evolution of vertebrates, while neutral, harmless mutations occur at sites that have varied much more during vertebrate evolution:
The consistency and predictability of genome-wide patterns of disease- associated and neutral amino acid variants reported here underscores the importance of the consideration of
evolutionary rates of amino acid positions in clinical and population genetic analyses aimed at understanding the nature and fate of disease-associated and neutral population variation.