Site Map  -->  Home  -->  Lab Philosophy

Opportunities and research philosophy in the Richardson Lab

"Analysis and Design of Protein Structure" (or "The Natural Philosophy of Proteins")


So why are we looking at proteins, why do we think we are particularly effective, and what sort of training do people get in our lab?

Our overall laboratory goal is the Protein Folding Problem -- why is it that a particular natural sequence of amino acids self-assembles into 1 or 2 unique 3D structures with very specific functions, controls, and interactions? Proteins are the simplest and closest to being understood of biologically relevant self-assembly systems. Self-assembly occurs at many levels: molecules, organelles, cells, tissues, organisms -- with very definite targets, limitations, and disastrous failures. Self-assembly is basic to understanding mad cow disease, Alzheimers, cancer, etc., as well as to predicting structures. We are obviously doing very basic research rather than applied research on a specific biological problem, but we find it very satisfying to know that our results have been and will be very broadly useful.

Our approach is studying the known structures to understand the principles and details of why proteins fold to specific patterns, and using that expertise explicitly to, for instance, design new ones and improve the accuracy in knowing old ones. Computer graphics and computing power are essential tools -- but we resist using computing power to compensate for understanding and coach our people in the same discipline. Much of our current work centers around a new approach our lab developed that adds and optimizes hydrogens and uses all-atom contacts to visualize and quantify the details of interactions within and between molecules. This new information is largely independent of other techniques, and it allows us both to see new patterns in structures and to tackle a new set of problems.

Because we are a small shop with two faculty members, group dynamics are unusual. Everybody is involved to some extent with everything going on in the lab. People do have special responsibilities for certain projects (and undergrad assistants for special projects are highly valued by grad students as well as post-docs). As well as whole-lab group meetings, there are frequent informal sessions; both are very productive and educational ways to work through questions and issues as they arise. Of course, sometimes people just need to be left alone to chug through something themselves. When two PIs with different talents are intimately involved with the day-to-day research, direct interaction with the PIs can be extensive (and very highly valued, by the PIs at least!).

Just as we welcome people from other labs to work with us on graphics projects, our people can learn specific skills or use instruments in other labs.

Our singular effectiveness stems, in large measure, from treating research as "data-driven" rather than "hypothesis- driven". We study the data with general goals in mind and allow hypotheses to arise from the patterns -- then, of course, we follow the basic paradigm of science and test these hypotheses against more data. Successfully following the "hypothesis-driven" paradigm seems to depend on developing the instincts to propose good hypotheses based on a preliminary look at the data and on the willingness to change or abandon them when appropriate. Learning to do this - that is, acquiring and honing those instincts - is explicit in our lab, a natural by-product of the way we work with all those who work with us.

Since our work is "data-driven", new ideas and possibilities continually arise. On the conviction that we will make the most significant contribution to science doing what interests us the most, we try to follow out such leads that have a good chance of being productive. For instance, our protein packing studies have led us toward developing methodology to make protein crystallography refinement faster, more accurate, and possibly fully automated. Along the way, we are continuing to develop even more powerful structure-validation tools.

The lab approach is to be very persistently and deeply critical not only of our own hypotheses, but also of our actual observations and methods. Our people learn how to be persistently questioning and still get things done, and to match the nature of the question to the precision of method and available data. The real excitement is the continual revelation of new and unanticipated directions to pursue that arise from the synergy of people of various talents working together.

Another very important philosophical aspect of how we do research and training is that we find "state-of-the-art" to be equivalent to "status quo", and not a place we find very interesting. We are always looking beyond state-of-the-art for the next cutting edge where we can do something that otherwise would not be done. We are developing methodology that lets one see finer details in protein structure than was heretofore possible. People in our laboratory thus get experience developing the cutting edges that allow science to progress beyond current state-of-the-art, as they concurrently gain a very intimate and useful familiarity with protein structure.

Just as state-of-the-art data collection equipment and computer programs were once inadequate to even solve protein crystal structures, we have discovered that current computer programs for structure refinement do not capture all the exquisite detail inherent in actual proteins. Our small-probe all-atom contact methodology (PROBE) adds a cutting edge to that process. Certainly the details of protein structures must be consistent with physics, but we see tighter distributions than predicted by current physical parameters and algorthms. We are working (with others) to reconcile the precision seen in validated accurate structures with molecular dynamics and quantum mechanics.

In spite of the great overall precision in high-resolution structures, macromolecular crystallography is not yet a fully mature art since different labs working with identical protein samples with the same conditions produce structure coordinates with significant differences in a few, often crucial, conformations. Those problem areas are independently found and can usually be corrected by the all-atom contact tools we are developing. We have a project using those tools to produce a corrected "consensus" structure that will fit each lab's diffraction data as well or better than their original coordinates. Concurrently, we are re-refining and correcting our old superoxide dismutase and Staph. nuclease structures along with newer data from other labs, to produce coordinate sets that are completely free of steric violations. These same improvement strategies, once tested and streamlined, will apply to individual structure determinations as well. Similarly, it should be possible to make a "chimera" of NMR models to get a true "best" structure. If indeed we can ensure protein models to be as accurate as the precision of the data -- then homology modeling, rational drug design, and even true, from-scratch de novo design may really become accurate enough to be, at last, useful at the level of function!

This "data-driven" but simultaneously data-critical approach is one reason our side-chain rotamer library is better than any previous one, and gives us confidence that we can describe a similar library of mainchain conformers. The first step in that process will be to characterize the range of Calpha geometry and of Calpha-Cbeta direction, distinguishing real variation in phi, psi and in bond angles from errors. In the end, careful selection and validation of the rapidly-growing structural biology data will lead to better understanding of the determinants of protein structure. The ultimate test of such understanding is de-novo protein design, heretofore a fascinating but frustrating endeavor. Re-design, however, has become successful and very useful, and we are working on redesigns of thioredoxin and a symmetrical version of ROP protein, as well as variants of lambda repressor, to test hypotheses about specificity (or uniqueness) of structure.

Besides adding blades to make a cutting edge, sometimes in sharpening a tool one applies the metaphorical grindstone to remove extraneous material, making a keener instrument. Our computer graphics program MAGE draws on 30 years of experience to use only those representations that make kinemages singularly effective to study 3D relationships. Interactively modeling mutations with MAGE and PROBE lets one easily decide whether a mutation is compatible with the known starting structure and therefore whether or not the results would be straightforwardly interpretable. We believe that both research and education are community efforts, and all of our programs and datasets are made freely available on our Web site. It is a goal with us to make our techniques and software easily usable by other people. As well as research instruments, kinemages are powerful educational tools. Our people serve as an informal research resource and teaching laboratory. Great fun, and a great opportunity for us to keep our minds on the broader picture.



[Home] Kinemages Java Mage Software: [MolProbity] Validation Rotamers FTP 2D Gallery Lab Info [Site Map]
Display Contacts Utility