One day, the boss was unavoidably elsewhere, so she missed the weekly seminar. When she came back in the afternoon, one of the graduate students and I were crowing about the best, the most inspiring, the niftiest ("Gad Whistler, I wish I'd said that") talk we'd heard that year; perhaps the best that decade. The presenter was Aled Edwards from the University of Toronto and he talked about a brilliant example of Harry Potter winner-take-all that is costing science dear. The Toronto hypothesis was that, after spending $1billion on sequencing the human genome, we should have something to show for it - new genes revealed, new associations of disease with mutations in the DNA, new drug targets, new ways to seeing. But it wasn't like that! Edwards & Co, analysed two generations worth of scientific literature to count up the number of published papers about specific genes of pharma-medico-biological interest. In particular they looked at druggable targets in three classes: kinases, ion-channels and nuclear receptors, before and after the delivery of the human genome sequence into the public domain by Bill Clinton and Tony Blair (thanks boys!). That's a lot of data to trawl through, as more than 20 million papers have been published since PubMed indexing started in 1950.
What they found, for all three rather different sorts of protein, was that we are still squeezing the last drops of info out of the same proteins that we were studying before the human genome became available. This despite a lot of objective data to show that there were lots of very interesting disease-associations and potential therapeutic targets in the "dark-matter" of the genomic universe. The scientific community is still doing same-old same-old on the tried-and-true. Some time ago, Nobelist Roger Kornberg claimed that scientists have a tendency to "fondle their problems" or to use an even less appetising metaphor, to pick at the scabs of their previous research rather than coming up for a breath of fresh air and trying a new avenue to investigate.
Here's one we did earlier with my Yr3 bioinformatics class. It is a tally of the number of papers published about 4 of the 10 Toll-like receptors that we humans have to detect the hostiles in the environment so that we can mount an appropriate immune response:
|TLR||PubMed papers||PubMed reviews|
There are three good (i.e. bad) reasons for this scab-picking.
1) It is easier for editors and referees to accept papers that are firmly in the context of existing research - critiquing earlier work or building on previous studies or doing the same analysis in a different species or different human population or different cell line.
2) Research funding bodies are chronically risk-adverse - they don't do Blue Skies and minority interest ERRa is blue skies compared to over-analysed diminishing-returns Era.
3) Research depends upon kit, and there are chemical probes available for the Harry Potter proteins but it costs person-years and multi-dollars to develop new probes for the emerging proteins. The regrettable upshot is that these proteins don't emerge because nobody will risk (investing in) investigating their functions and attributes.
I thought that analysis was brilliant. Like the failures in ethics, failures in statistical power and failures of reproducibility, fondling is systemic in science (blobolinks). In the Nature commentary that sums up this research, Edwards offers some suggestions about how we might break free of these incestuous shackles of familiarity. My Institution is too poor to afford an electronic subscription to Nature, so I can't read that on-line but you may be able to. A longer version of the same material is defo in the public domain. This sort of research is also quite empowering because it shows you that useful and informative research can be done without horsing through thousands of dollars worth of kit and consumables.
Post a Comment