Jules
27th January 2003, 11:49 AM
Quoted.....Because I'm too tired to think for myself ;)
01/22/2003
- We are fast approaching 10 billion again, and have a 6.5A structure! So congratulations to all. This particular protein was expected to get roughly a 6A structure in a sample of 10 billion so we have come very close to that prediction. We will be changing proteins again next Tuesday afternoon, EST. Also coming soon, a brand new smarter algorithm. Watch here and in the discussion forum for details.
- Results from the CASP5 structure prediction competition have been posted in the results section, with a brief explanation of how to interpret the graphs. In most cases our predictions were in the middle of the pack. This might not sound as good as we had hoped, but it means none of our predictions were THAT bad, and in general all five of our top predictions were of comparable quality, meaning that we were picking out structures consistently. It is also important to keep in mind that of the 13 targets predicted, only T0129, T0139, T0161 and T0181 were truly 'new folds'. Other targets could in some cases be predicted by other means starting from related known structures (though still by no means easy to predict). We still need to go through the structures we generated to see if we submitted the best predictions we could have, or if there were better structures that our scoring function simply failed to pick out. Clearly brute force alone is not sufficient to beat other more intelligent search methods though, and we will shortly be revising the algorithm to work in a more intelligent, iterative sampling approach, which we expect to have a huge effect on the overall performance and efficiency of the sampling.
Have a great Monday :sunshine:
... :yawn: Going back to bed for an hour
01/22/2003
- We are fast approaching 10 billion again, and have a 6.5A structure! So congratulations to all. This particular protein was expected to get roughly a 6A structure in a sample of 10 billion so we have come very close to that prediction. We will be changing proteins again next Tuesday afternoon, EST. Also coming soon, a brand new smarter algorithm. Watch here and in the discussion forum for details.
- Results from the CASP5 structure prediction competition have been posted in the results section, with a brief explanation of how to interpret the graphs. In most cases our predictions were in the middle of the pack. This might not sound as good as we had hoped, but it means none of our predictions were THAT bad, and in general all five of our top predictions were of comparable quality, meaning that we were picking out structures consistently. It is also important to keep in mind that of the 13 targets predicted, only T0129, T0139, T0161 and T0181 were truly 'new folds'. Other targets could in some cases be predicted by other means starting from related known structures (though still by no means easy to predict). We still need to go through the structures we generated to see if we submitted the best predictions we could have, or if there were better structures that our scoring function simply failed to pick out. Clearly brute force alone is not sufficient to beat other more intelligent search methods though, and we will shortly be revising the algorithm to work in a more intelligent, iterative sampling approach, which we expect to have a huge effect on the overall performance and efficiency of the sampling.
Have a great Monday :sunshine:
... :yawn: Going back to bed for an hour