"So, what are you studying?" Depending on who is asking, the length of my answer would vary. See, since I go to grad school, quite a number of people are asking me that; and answering 'Biomolecular NMR' would draw some blank stares.
So, go back a little: Chemistry. Ah, my undergrad major, and since my lab uses analytical chemistry technique (i.e. Nuclear Magnetic Resonance) to probe biomolecular structures, this answer is not too far off the mark.
"So why are you in the School of Biology?" Ahem. Here we go again.
So, the slightly longer and my supposed official areas of study: Structural and Computational Biology. I think I shall reintroduce my clockwork analogy here. Imagine a clockwork with all its gears running inside. If one would want to know how exactly the clockwork mechanism work, one would want to take it apart and see the gears; how they are connected together, what are their shapes that fit each other so that the concerted mechanism is running. Now, every living thing is also a machine, a biochemical machinery, that, like a clockwork, consists of tiny, tiny parts that are grinding away and give rise to life as we know it. Ideally we want to map how these biomolecular gears interact with each other (i.e. the interactome). Prior to that we would want to know how these tiny gears look like. The structures of these tiny gears are then what the structural biologist would like to know.
'To know' here, as in all areas of science (or anything, really), should be interpreted philosophically. The usual limit of what a structural biologist want to know goes beyond 'having a visual representation'. Sure, it's nice to know what these biomolecules look like since we normally can't see them with naked eyes. But blobby blobs at crappy resolution (say, 15 Å) doesn't tell much. Like a blurry text on a document for example; it's unreadable. What we would like to know, ideally, is the visual representation at the resolution of individual atoms (i.e. atomistic resolution). The blurry text on your document now gains some sharpness; now you can distinguish individual letters; it's now legible. Often, this is sufficient to establish how it behaves, how it interacts with, say, water molecules, metal ions, drug molecules, or other biomolecules.
To even arrive at this point is an accomplishment (to illustrate, resolving a previously unresolved structure of a big protein and analysing its structure-function relationship may occupy the whole PhD thesis, to give you a sense of scale). Could we do better, you ask? Erm, there's actually some problems -- hitting the philosophical wall, so to speak. Let me explain.
*WARNING: TECHNICAL TERMS AHEAD*
So, there are two common techniques used to probe biomolecular structures. (Oh, when we are talking about biomolecules, we mostly talk about proteins -- the biomachinery mostly comprises them).
The first is X-ray diffraction. It works by interpreting the diffraction pattern of X-ray by electron density of the atoms that make up the biomolecule. Since the atoms are arranged in a particular way for a biomolecule, the diffraction patterns are unique to each biomolecule; and long story short, the X-ray crystallographer would be able to construct back the electron densities of the biomolecule, thus she would able build a model of the biomolecule.
Now, problem #1, for a nice diffraction pattern, you would need repeating units, so you would need crystal of your protein, and crystallising a protein is honestly a PITA. Arising from that, problem #2, you would try a lot of different conditions until you find one that is suitable for crystallisation -- who is to say that that particular condition doesn't affect the structure of your biomolecule? Still arising from problem #1, the protein in say, your body, is in aqueous environment, but in your crystal it is obviously in solid -- problem #3, who is to say that your protein structure is not altered by crystal packing, and problem #4, your final structure would be a static snapshot of the structure frozen in the crystal, while the structure in solution must be a lot more wiggly-diggly -- do you now claim you 'know' the structure? (There, your epistemological wall). Problem #5, most of the atoms comprising a biomolecule -- C, O, N, S, Se, and some metals -- have nice, fat electron densities, but that ubiquitous Hydrogen, the most abundant element in the Universe of things living and non-living, does not. As such, H atoms are not visible in the crystal structure. Problem #6, so are regions of the biomolecule that does not form regular repeating units (most likely because they are floppy). Problem #7, when you are matching atoms to your electron density map, some are ambiguous (e.g. you can point sidechain A at this orientation, it fits ok, flip sidechain A 180° and it still fit ok, even though on one orientation it's N but the other is O -- since N and O have similar electron densities).
So your typical X-ray structure file would miss its H's and its flexible regions (the termini are usually susceptible); some sidechains may not be at their correct orientation; and there's only one frigging, static frame. Would you now consider you 'know' the structure?
Oh, then there's NMR. The laypeople would be more familiar with MRI (Magnetic Resonance Imaging). Indeed the underlying principle is the same: both exploit the same phenomenon, i.e. nuclear magnetic resonance which arises from nuclear spin (which belongs to the wonderful, you-will-understand-better-if-you-don't-try-to-understand-it world of quantum mechanics (for instance, there's no way to visualise quantum spin)). Long story short, NMR yields a set of restraints. If in X-ray crystallography one needs to fit the biomolecule to the little pockets of electron density, in NMR one needs to fit the molecule to satisfy these restraints. So you would have to know the amino acid sequence of your protein. So then you have a totally linear peptide -- make the peptide explore different conformations to satisfy the restraints, a game of Twister if you will. What you end with would be a set of conformers, which would be ranked according to their energy. Usually the first 20 low-energy conformers are deposited in public domain. The fact that there is a set of conformers, called an ensemble, is important. By aligning the conformers in an ensemble, it's usually clear which region is flexible -- the deviation for that particular region would be more pronounced across the conformers. So an NMR structure would have its H's, all amino acids visible and it's not a static snapshot (cf. X-ray structure). Also, if you do solution NMR, it's may be possible to as close to physiological condition as possible (may not be totally tractable because of limitation in NMR setup itself, e.g. NMR signal at high pH condition decays too fast).
While seemingly answering to all X-ray crystallography's woes, the NMR spectroscopist faces a totally different set of problems. First off #1, NMR is an indirect technique, unlike X-ray diffraction which directly constructs a map of electron densities. So NMR structure doesn't really have a concept of 'resolution', which makes it kind of hard to assess the quality of a structure. So, you have made your protein play a game of Twister to satisfy your restraints, now #2, the quality of restraints -- if I see a flexible region in the ensemble, is it really flexible region, or I didn't apply enough restraints, that's why it's fuzzy? #3, the game of Twister is called restrained molecular dynamics, basically a computer simulation. Of course, it's only possible to incorporate approximations and assumptions, not the whole physical law shebang, into the software. Some of those may not be appropriate and may affect the resulting structure in a significant way. There are some technical obstacles as well, like how the protein of interest needs to be isotopically labelled with NMR-active nuclei (15N, 13C). The final step of reconstructing the structure is tedious both for X-ray crystallography and NMR spectroscopy -- so that adds to the time needed to produce a good 3D representation of a protein.
So, for the umpteenth time, after successfully determining a structure, the structural biologist should ask, metaphysically, Do I 'know' the structure?
Science, people.
That's it. Next time I will just reply, "I do Science."
