PAE and other matrix metrics will expand [quadratically by protein length. This can take up a lot of space (100s MB per structure) but not as bad as .pkls
Despite the downstream demand of reading through text-formatted modelCIF PAE into something like ChimeraX, my thought is fully embed in the modelCIF file, use .gz compression to get whatever savings we can. Rather that having to pass multiple files around and deal with viz program issues.
It would be far more efficient to use numerical datatypes (.npz), so that is more computationally efficient. Are the performance gains better than the trade-off?
PAE and other matrix metrics will expand [quadratically by protein length. This can take up a lot of space (100s MB per structure) but not as bad as .pkls
Despite the downstream demand of reading through text-formatted modelCIF PAE into something like ChimeraX, my thought is fully embed in the modelCIF file, use
.gzcompression to get whatever savings we can. Rather that having to pass multiple files around and deal with viz program issues.It would be far more efficient to use numerical datatypes (
.npz), so that is more computationally efficient. Are the performance gains better than the trade-off?