A recent issue of Nature (443, 382, 28 September 2006, subscriber link) carried an intruiging article about Japan's five-year "Protein 3000" project, which is now winding down. Carried out under the auspices of RIKEN, the project was designed to use a large-scale NMR facility to solve the structures of at least 3000 proteins, and along the way advance the understanding of protein folding in solution.
Whether or not it succeeded depends on who you ask, because the answer isn't obvious. The project does seem to be on track to make its numerical goals, but according to the article, many protein-structure people think that a large number of the structures that have been solved are, well, junk - easy, closely-related ones that were put on the list to run up the numbers. While the organizers dispute that, as they certainly would, another problem is that understanding protein folding has turned out to be (you know what's coming) harder than expected. The project was supposed to cover a large swath of a hypothetical 10,000 different folds, but now the real number is thought to be two or three times that. So the best case was that Protein 3000 would have worked out about a third of all possible protein folds, but now they're looking at perhaps 5 to 10% of them.
The Japanese government has a real weakness for big programs like this. I think that Protein 3000 has been one of their biggest forays into that area, but in the past they've announced all sorts of gaudy projects in computation and the like, most of which haven't worked out quite as planned. The "Fifth Generation" project is perhaps the most abject failure of the lot, but at least that one seems to have produced a number of researchers who could do something else. But the Protein 3000 business has some folks worried:
Several researchers have also expressed concern that the factory approach at the NMR facility has deprived young researchers there of the skills necessary to solve more complicated and important scientific riddles. It might have "destroyed the next generation", says one.
(Kurt) Wüthrich, who helped plan the NMR centre in 1998 and was a science adviser in 2000-04, agrees that the facility is a wasted opportunity. "A centre of that size should contribute to methodology, but there has been nothing," he says. "It became a one-man show with 40 NMR machines - there is no knowledge."
Not a good review, considering it comes from a man who knows a bit about the use of NMR to attack protein structure. What I find instructive about such things is that these projects are often just the sort that large government-level granting agencies take it into their heads to fund. Sometimes they work out, but the majority of the time they don't.
1. Chrispy on October 9, 2006 9:52 PM writes...
Permalink to CommentDo you know if they have released the coordinates of the proteins they have solved? RIKEN has historically been very secretive.
2. Tom Womack on October 10, 2006 2:20 AM writes...
Hasn't there been at least one equally large-scale privately-funded embarrassing failure recently - Paul Allen's mouse-brain maps which demonstrated that most proteins are either not expressed in mouse brain or expressed uniformly across mouse brain?
Large-scale structural proteomics by crystallography strikes me as maybe an equally peculiar field: there seem to be projects which are running down a sequenced genome looking for reading frames, running the reading frame in E. Coli, and, if the protein happens to dissolve and happens then to crystallise usably, running the diffraction process and submitting to PDB. Meaning that you end up with lots of structures for proteins of unknown function, annotated at best with 'homologous to known-protein', and what you're looking at is the subset of the proteome which crystallises nicely, which isn't self-evidently biochemically sensible.
It does mean that people have got round to robotising the process of growing protein crystals; it remains a black art followed by an exhaustive optimisation, but at least it's now one carried out 24/7 by machinery rather than by grad students. Which allows for one of the newer small-pharma concepts of crystallising a known-interesting protein in bulk, soaking the crystals in the compounds from a whole library, and doing crystallography to see which ligands bind where. Not sure that's produced any drugs, or even any decent leads, yet ... it gets you candidates which bind well to useful things, but no specificity, so I'd have thought you end up with a handful of leads among a multitude of highly-toxic general enzyme inhibitors.
Permalink to Comment3. highlyreactive on October 10, 2006 5:22 AM writes...
"While the organizers dispute that, as they certainly would, another problem is that understanding protein folding has turned out to be (you know what's coming) harder than expected."
It's interesting that you say that. Baker (at UW) seems to think that it is easy to understand protein folding. At least that is what he seemed to be saying in an article in Nature quite a while ago. I thought that was perhaps a highly optimistic appraisal even with his successes in CASP.
Permalink to Comment4. Derek Lowe on October 10, 2006 8:03 AM writes...
I have a Mr. Spock raised-eyebrow expression ready for anyone who tries to tell me that the protein folding problem is easy.
Permalink to Comment5. Ashutosh on October 10, 2006 1:33 PM writes...
Protein folding may never be 'solved' but heuristic and knowledge based approaches combined with first principles physics based approaches may make the problem tractable on a phenomenological basis. Baker has really done some remarkable work. His Science TOP protein de novo design paper and the recent Nature endonuclease design studies are hailed by many to be definitive.
Permalink to Comment6. Hap76 on October 10, 2006 2:26 PM writes...
This sounds a lot like the tendency of the (presumed) tendency of agencies funding foreign aid to give towards big projects which allow them and their donors to get their name on things but which mitigate the problems they propose to solve very little, because the followup was poor and because the donors needed a large project more than the (supposed) beneficiaries of the aid actually did. Might something similar be going on in the creation of big name projects?
Permalink to Comment7. srp on October 10, 2006 6:51 PM writes...
Hap76, you might want to look at Freeman Dyson's essay in From Eros to Gaia on big "Plan A" projects that absorb a whole field's resources and more modest "Plan B" projects that solve narrower problems more incrementally and increase skill levels. It's really amusing and brilliant, with lots of good examples.
Permalink to Comment