Sunday, February 22, 2009

From human genetics to biological insight   posted by p-ter @ 2/22/2009 11:20:00 AM

In 2007, SNPs in an intron of the gene FTO were reported to be associated with obesity. At the time, essentially nothing was known about the gene. A few months later, a group of biochemists proposed a role for the gene in demethylation of nucleic acids (RNA or DNA). This week, a group of mouse geneticists present an analysis of a knockout of the gene, and show that the knockouts are resistant to weight gain due to increased energy expenditure.

There's still quite a ways to go before the mechanism by which FTO contributes to weight variation in humans is understood (oddly enough, there's some evidence that the mechanism is through increased energy intake rather than expenditure), but people keep chipping away...

Labels: ,

Thursday, May 17, 2007

Arrays on the way out?   posted by p-ter @ 5/17/2007 09:57:00 PM

A few months ago, I wondered out loud if next-generation sequencing techniques would render microarrays obsolete. There's no need to estimate abundances of any nucleic acid by hybridization if you can just count up frequences of mRNAs or DNA or whatever directly via sequencing. I must have heard that idea somewhere, because this week's Cell has an example of just that principle.

The assay in question here is chromatin immunoprecipitation (ChIP), a classic technique for testing whether your protein of interest is bound to a DNA sequence of interest. In the assay, you crosslink proteins to the DNA, then use an antibody to your favorite protein to isolate the DNA it's bound to. In the old days (or now, if you have a specific hypothesis to test), you would then do PCR on said DNA to test for the presence of a sequence of interest. With arrays, you could blindly hybridize the DNA to the array, generating a map of where your protein bound in the genome (in this case, the assay is cutely called ChIP-chip). But this is subject to all the problems of arrays-- isolating the signal from the noise, possible low resolution, a need to amplify the DNA, etc. One way around this is to simply sequence the DNA-- binding sites can then be directly imputed from the sequence data (the quality of which will depend on the efficiency of the IP). This is exactly what the authors do, and it seems to work.

Moral of the story: technology moves fast.