I have regularly been struck by how closely related proteins crystallize in completely different conditions. Eg see Galina Obmolova and colleagues' excellent paper (ref below). A set of sixteen highly homologous Fabs were crystallized in apparently random conditions. Roughly half used PEG, half high-salt conditions, salts used included sulfate (sic!), formate, citrate, acetate, tartrate, chloride etc etc, while pHs ranged from 4.5 to 9.5.
I also once downloaded the entire PDB and looked for correlations between the reported crystallization conditions and the (summed) areas of every atom of every residue on the surface of the proteins. I came up with . . . hmm . . absolutely nothing useful. I conclude that we need to start with regular screening for pretty much every new sample that we have - so an unbiased empirical approach is the only good way to go. Thx Patrick Galina ref - Obmolova, G., et al. "Protein crystallization with microseed matrix screening: application to human germline antibody Fabs." *Acta Crystallographica Section F: Structural Biology Communications* 70.8 (2014): 1107-1115. On Tue, 23 Jul 2019, 23:53 Newman, Janet (Manufacturing, Parkville), <janet.new...@csiro.au> wrote: > There are a bunch of people doing this – in the small molecule world. And > a lot of work has been done on some very robust protein systems too. Can > you guess which ones? > > > > The real issue (at the moment) is that all the pre-work needed to predict > if or how a protein might crystallise takes more work and more protein than > setting up crystallisation experiments. > > How many people do DSL on protein in a crystallisation screen, for > example? Or do self-association chromatography to determine the B22 (which > changes under different conditions, naturally). Or try mapping out a phase > diagram (for each condition)? > > > > Many people are not even aware that a simple PCT can help one work out a > sensible starting concentration for crystallisation trials. > > > > As for AI, at the moment unsupervised learning doesn’t seem to do much, > which means we need vast, well annotated datasets to make progress. MARKO, > which Sarah mentioned, required half a million scored images, which took > years to get together. > > > > Janet > > > > *From:* CCP4 bulletin board <CCP4BB@JISCMAIL.AC.UK> *On Behalf Of *Keller, > Jacob > *Sent:* Wednesday, 24 July 2019 4:18 AM > *To:* CCP4BB@JISCMAIL.AC.UK > *Subject:* Re: [ccp4bb] challenges in structural biology > > > > What about developing a theory of how crystallization happens, i.e., what > does the microscopic “picture” look like when crystals are forming, then > predicting based on that picture? I remember looking into these things > about ten years ago, and there were some cool things being done with > various scattering methods and with AFM, but am not sure now what is the > state of that art. > > > > It would seem to me that crystallization is the search for intermolecular > docking sites of sufficiently good (albeit presumably weak) affinity and > consistent with the formation of a 3D lattice. I wonder what the affinity > of these sites is, actually—I guess somewhere in the micromolar range, > based on usual protein concentrations under crystallization conditions (10 > mg/ml of a 40 kD protein is 250 uM). > > > > Presumably the various docking sites would change affinity based on the > crystallization conditions, which would explain why some crystallization > conditions work, others don’t? > > > > Maybe a systematic look at all crystallization contacts in the PDB might > yield some insight into crystallization? Maybe it’s already been done? > > > > JPK > > > ------------------------------ > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1 > > ------------------------------ > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1 > ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
