...some more updates and progress reports that David Baker posted in the Rosetta@home forum over the weekend - thought I copy them over here in case you missed any of this:

HIV vaccine design:

so far 12 vaccine designs have passed the first tests: we can make them in large quantities, and they interact tightly with antibodies known to neutralize the HIV virus. These designs are now being tested by our collaborators to determine whether they elicit a neutralizing response to the virus. We expect that this will not be achieved immediately and we are currently generating another round of designs. We will describe this work at a meeting of the Gates foundation in Seattle this coming Monday-Wednesday which brings together the different HIV vaccine design efforts they are supporting. I am looking forward to discussing ways in which the computations you are helping us with can contribute to all of these vaccine design efforts.

protein-protein docking:

The most recent illustration was just yesterday when Chu presented the results of your protein-protein docking calculations at our weekly lab seminar. They are really amazing--he showed that for the first time it is possible to allow full flexibility of both protein backbones and dramatically improve the accuracy with which the structures of protein-protein complexes can be predicted. This simply would not have been achieved without the large scale computing boinc and all of you are making possible as the space of possibilities is FAR too big to be searched with conventional computing resources. The very striking improvements in protein structure prediction evident in the CASP7 tests (see my journal) are another very recent example.

CASP7:

The viewgraphs from the CASP7 conference of last week have now been placed on the web. This includes presentations by Bin Qian and Rhiju Das from the Baker Lab, explaining their algorithms and showing lots of examples (their last viewgraph acknowledges the contributions of 100,000 Rosetta@home participants and shows a screenshot of the homepage) and by the CASP7 'assessors' doing a detailed comparison of the quality of the predictions by the different groups. The Baker Lab is one of the top two groups that stand out, the other being Yang Zhang from the University of Kansas. David Baker mentioned that he had already invited Zhang to Seattle and that datasets and code are being exchanged to learn from each others approach:

There are two sources of information which protein structure prediction can draw on: evolution and physics. Zhang's approach is excellent at using all evolutionary information available from related protein structures, wheras our search for the global energy minimum uses physical chemistry information. As you suggest, the two approaches are quite complementary and we can each learn from each other. I invited Zhang to the University of Washington last month for a seminar, and we spent an excellent day discussing routes for collaborating, and we are currently exchanging datasets and programs. Using evolutionary information will help us narrow down the part of the space that needs to be searched so that we can better focus your searches on areas where the global energy minimum (correct structure) is likely to be, and thus improve prediction of larger more complex protein structures.