![[syndicated profile]](https://www.dreamwidth.org/img/silk/identity/feed.png){kind=small}
in_the_pipeline_feed) wrote2025-08-13 04:40 pm
The Frontiers of Structure
Let’s have a look up near the state of the art in structure determination for challenging samples (small amounts, difficult-to-impossible crystallizations, and so on). This area has been advancing a lot in recent years, especially due to improvements in electron diffraction and cryo-electron microscopy techniques. We keep gaining more and more ability to see things that we never really could before.
For example, here’s a recent preprint from a team of Korean researchers who are pushing the size limits of cryo-EM. This 1995 paper is the standard reference for what those limits would be, and the one that’s really hard to escape is radiation damage. Anything that’s got enough oomph to go through/past/off of a sample and provide structural information afterwards is going to have enough of that oomph to start degrading it. Electron-based methods offer the best information-per-damage ratio, definitely less strenuous than X-rays. Neutrons can provide pretty interesting data, and the (rather heavy) damage associated with them can be partially mitigated by switching as much as possible to isotopes like deuterium that have a lower neutron capture cross section. But the world lacks any really bright-and-tight neutron sources as compared to what we can accomplish with electrons and X-rays.
At any rate, the estimate for cryo-EM is that you can work down to molecular weights of about 38,000, but going lower than that will be tough. It gets harder and harder to fit the data by just looking at particle after particle in all kinds of orientations without any underlying crystalline order - I suppose that advances in computation might help, but my impression is that so far they’ve mainly helped to realize such structure-solving in shorter and shorter times (and in a more and more automated fashion) without necessarily pushing the size limits much.
The record until now was 46,000, but this group pushes that down to a 2.32 Å structure of maltose-binding protein at 43 kDa, and a 3.04 Å structure of a human PLK-1 domain at 37 kDa. The former is good enough to spot the maltose in there and a network of associated water molecules, and the latter, while lower-resolution, is still good enough to model a bound small molecule ligand (onvansertib). I would assume that that 38 kDa limit has some error bars on it, but this is still very good to see. As the authors point out, RNA structures could be expected to go even further, because all those P atoms scatter electrons well. Watch for this limbo bar to be set even lower!
Over on the micro-electron diffraction side of things, this preprint from SUNY-Buffalo shows a very high-resolution structure obtained from spontaneously formed protein microcrystals of the protein crambin. These came straight from a drop of ethanolic solution during purification, but proved to be pretty useless for X-ray diffraction because of their size. But by shooting several dozen such crystals with a microED rig, the combined data set yielded a structure at 0.85 Å, which is mighty fine for a protein. Crambin has been a real proving ground for structure determination - x-ray synchrotron data have yielded a 0.45 Å structure, and neutron diffraction a 0.54 Å one.
The authors note, though, that the great majority of microED structures have had a leg up: similar proteins or AlphaFold structures are used to get a handle on the solution from the data. Here, though, they wanted to see if they could solve this one ab initio without reference to anything save a five-residue helical fragment. You’re not always going to have a homologous structure that someone has already worked out, and AlphaFold (and similar software) will choke on proteins with unusual folds that they haven’t been trained on.
Interestingly, the team found that (as mentioned) the spontaneously formed microcrystals weren’t worth as much in X-ray, not least because of rapid radiation-induced degradation. They could grow larger conventional crambin crystals which were just fine for XRD, but these were too thick for microED. They then crushed these into smaller pieces, hoping to produce a whole array of useful small crystals for electron work, but these turned out to be notably inferior to the ones that just fell out of the ethanol droplets (!) That led to the “serial crystallography” approach with multiple examples of the spontaneous crystals in combination.
There were of course challenges - plate or needle-like crystals suffer from gaps in their data sets due to one or more thin dimensions and the angles that these force the majority of data to be collected at. But the authors show a useful workflow to mitigate this problem and provide anisotropy-corrected data (without introducing artifacts). And the fragment-based solution to the phasing problem is also usefully detailed and will be of interest to practitioners. The hope is that this will serve as a benchmark for the next generation of electron-based equipment and methods.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
![[community profile]](https://www.dreamwidth.org/img/silk/identity/community.png){kind=small}
