David Bogdanoff wrote (15 Jun 1998):
"... He (Hazeltine) said that most human genes had been identified (not to be confused with the most of the human DNA material whose function is largely unknown). If this is so, can you please enlighten me as to what AFFX chip is supposed to do? Apparently, it is not to identify large numbers of unknown human genes, since these have been identified for the most part according to Hazeltine. Thanks in advance for your response."
First off, for those who'd like a bit of technical background on the isolation & sequencing of human cDNA, there is the March 1997 Scientific American article by Hazeltine (pp. 92-97).
Second, be careful by how you define "identified". There is the caveat that, although one might successfully sequence a cDNA (knowing thus that it represents a transcribed human gene sequence; sometimes the partial cDNAs are called Expressed Sequence Tags, or ESTs) - you might not know what *function* it serves unless (a) you can identify substantial sequence similarities between the unknown cDNA and a DNA sequence of known function, or (b) you can demonstrate the function of the cDNA's gene product (either in a human cell, or some other cell like yeast where you rescue a function you know is missing).
But, if you *do* have the sequence(s) in your hot little hands, then you can perform novel chip-based analyses to gain understanding to the function of the sequence. Here's just two ways that have come out of published AFFX research:
1) you can monitor the expression level of the sequence(s). Is it made or not made in a certain cell or tissue type, at a certain time, under certain conditions/signals? If so, how much is made? With an AFFX chip, you can array out DNA probes complementary to a very large number of these cDNA/EST sequences (a few dozen probes per transcribed unit usually gives good confidence) so that you can determine the expression of these unknown genes *in parallel* in the hundreds or even thousands (it's that "in parallel" that gives you the great savings of effort, time & cost). Looking at the patterns of when, where & how much mRNA is expressed usually gives the researcher some clue as to the possible function of the gene sequence [read Lockhart et al (1996) Nature Biotechnology v14, pp1675-1680].
2) Roughly about once per 1000 bp of human DNA sequence, you're going to find a common base variation (or "polymorphisms") in the population (let's say base 'A' versus base 'B'). If you can monitor a very large comprehensive set of such variations (in particular, those concentrated in protein-coding regions), then you can compare phenotypic traits distributed in a human population sample against "who in the sample has two copies of variant 'A' versus who has two copies of variant 'B' (or even one of each)?" Thus, you can do genetic linkage analysis to determine which biological functions are linked along the chromosomes to which sequence variations, and an AFFX genotyping chip can allow you to look *in parallel* at thousands of such polymorphic genetic markers in a single experiment per individual [read Wang et al (1998) Science v280, pp1077-1082].
I hope this reply to your question doesn't sound too technical (I tried to pitch it at the level of Sci_Am; if someone can put it into simpler language, please do). But the basic gist of what I'm trying to say is that analyzing human cDNA/ESTs of unknown function (of the sort HGS hunts for) using various REsequencing-by-hybridization assays (as the AFFX chips do) is a highly parallel (and thus high-throughput, lower-effort, less-$$$) process: much more so than gel-based DNA sequencing. That chip results are shown to be highly-reproducible and consistent with conventional methods also means that the analyses are of high quality. Good + cheap + fast = high value!
-- Doctor Goodhybe |
