by Peter Nollert
December 23, 2009 02:47
In many high-throughput protein crystallization laboratories protein crystallization is dealt with as a standard process that usually goes like this:
1. Search for protein crystallization condition using a standard array of crystallization reagent matrices (BTW - many labs use Wizard I, II, III and IV for such a first pass ;). These experiments are of the trial-and-error type, sparse matrix screening that is akin to shooting in the dark. Protein crystallizaiton trials are set up with volumes as low as the liquid dispensing tools at hand allow and plates for sitting (and, less frequently) hanging drops in fairly standardized way (i.e. combining 300 uL protein solution + 300 uL precipitant). Plates are then stored in the dark at one or two temperatures and inspected after a day, several days and weeks. Once crystals or crystal-like objects are identified, the next step concerns
2. Optimizing the protein crystallization condition to grow better X-ray diffracting protein crystals. At this stage the repertoire of protein crystallization optimization procedures explodes, and one needs to consider available resources, experience and the protein at hand to apply them best. Popular crystallization optimization procedures include:
• Systematic grid screening
• Temperature variation
• Additive screening
• Seeding
The good news about these optimization procedures is that they are systematic and the results of the crystallization optimization experiments are often informative because they point to trends. For instance, you can identify a temperature that produces larger protein crystals. However, these optimization experiments need to be carried out in a fairly disciplined way, as one dimensional variations of a single parameter, keeping all other parameters constant. The alternative would be screening an astronomical number of crystallization conditions:
There are only a few points in the multi-dimensional protein crystallization space that one visits in typical protein crystallization experiments. Looks daunting, doesn't it?
And here's the problem with this scheme (not that I have a solution to it, but I'd like to point it out nevertheless):
By screening just one dimension at a time, large spaces of the multidimensional crystallization phase space remain unseen. This I call the Curse of Dimensionality in protein crystallization. I'm wondering if there are there any practical ways to getting around it. Any ideas?
There's no need to get too tripped up with this, though. The good news is that most proteins can be made to crystallize. In my mind this means that we have quite some slack in defining protein crystallization conditions.
Happy Holidays,
Peter