Seeing and knowing in structural biology
Carolyn Cohen, Prof. of Biology at Brandeis U., gave a talk with the title 'Seeing and Knowing in Structural Biology' at the 2000 Biophysical Society meeting in New Orleans. Her thesis in this paper was that 'seeing an image... often poses an enigma, and the solution of that enigma allows seeing to become knowing.' Dr. Cohen cited a number of examples from her own work and that of others, especially in the field of motility. One problem in this area is the ambiguity involved in assigning biochemical states in the contractile cycle for the three conformations of the myosin molecule which have been determined crystallographically. We quote here the coda of the talk, since we believe it expresses well one point of view toward current approaches to Structure/Function studies.
"...In my final remarks, I want to emphasize how we are, in fact, only at the threshold of knowing how the motor machine splits ATP to produce contraction. Of course, in order to explore what I shall call 'conformational space' for myosins, we need a thorough crystallographic study of a wide variety of isoforms to visualize the possible structural steps in the contractile cycle. Lee Sweeney's recent imaginative prediction about the polarity of myosin VI ... is a beautiful example of a new structure that will test and clarify our current thinking about the working of the myosin machine.1 ...[And] Yanagida's recent single molecule experiments. ... point to the need for a coordination of all of our current techniques in molecular biology, biophysics, and biochemistry to test and interpret functional models. Without a broad scientific culture, it seems unlikely that we can interpret an image correctly - or test our interpretations.
All of the work I've described today has followed the traditional path of 'determining the structures of proteins known to be biologically important.'2 And to someone like myself, the rate at which such structures are being generated seems very great. But, as many of you know, there are also plans to create so-called 'fast throughput' centers for protein crystallography to speed up this process, 'as in an assembly line.' And to generate, as well, a more comprehensive family of protein folds deduced by computer modeling from the great number of sequences now becoming available. There is the notion here that, with or without the aid of supercomputers like the IBM Blue Gene, these computational efforts will yield useful global structures for these proteins. On this logic, the next step - the one I’ve spoken of this morning - seems to some not to be a great hurdle: that is to proceed from either a detailed (or real) protein structure or from a global (or virtual) structure to an understanding of both its function and how it accomplishes this function. Some aspects of these enterprises may well turn out to be stimulating and useful - and some may be entertaining, as well. Certainly, the new conceptual and computational approaches, such as the use of phylogenetic profiles by Eisenberg and colleagues3 which discover functional links between different proteins, will be immensely revealing.
But I maintain my conviction that an understanding of functions will come - as it always has - from the creative insight of a few individuals. And that their understanding will be both inspired and constrained by the evidence for this solution. It is also plain that in many cases a detailed solution for one structure will disclose the ways in which other - more or less closely related structures - operate. I think that this route for discovery in science - this passage from seeing to knowing - has a great tradition whose lessons we should remember...or learn again."
 Wells, A.L., Lin, A.W., Chen, L., Safer, D., Cain, S.M., Hasson, T., Carragher, B.O., Milligan, R.A., Sweeney, H.L. Myosin VI is an actin-based motor that moves backward. Nature 401, 505-8 (Sept. 30, 1999).
 Thomas Steitz, cited in Wadman, M. U.S. plans giant effort on protein structure, Nature 400, 494 (Aug. 15, 1999).
 Eisenberg, D., et al. A combined algorithm for genome-wide prediction of protein function. Nature 402, 83-86 (Nov. 4, 1999).