who validates that the protein was folded correctly? Validating that the hash of some nonce contains enough leading zeroes is very easy for the network to validate. But verifying that users are correctly computing the protein folding (rather than quickly dumping out invalid folds to test hash values) seems like a difficult problem for the network to validate.
BOINC tasks are validated by running multiple executions on different hosts/users and comparing results; credit, and hence GRC, is only attributed to tasks with valid results: https://boinc.berkeley.edu/trac/wiki/CreditNew
This is a good point. I was assuming that computing a protein fold was something that could be done relatively quickly, but it seems like these simulations are actually much more computationally intensive than that.
This is obviously not my area of expertise, but I wonder if there is any way to make more efficient but still useful computational questions about protein folding. For example, if we try to fold just part of a protein (only 100 or so amino acids, say) and do it at a much lower fidelity of simulation, is the result still helpful?
