There are and may forever be mysteries of the atom but they all appear to be nuclear. As a chemistry teacher, I would like to know if there is anything left to know in the area of chemical reactions in terms of their predictability, modelling, energy states etc.
I guess, in other words, do we know enough about chemical reactions to accurately model them on a computer?
From a layman’s point of view: I’d have thought this was old hat by now.
This is more biochemistry than plain old regular chemistry, but the three dimensional folding of proteins is still exceedingly difficult to predict, which sucks, because the 3D shape is key to how they work. We’ve made a lot of progress over the years, but mostly through cheats and hacks. We’ve figured out that a lot of proteins are modular, so if we can crystallize and observe one example of each module, we can assume that other versions of the module are more or less the same, for instance. The fundamental question of how to go from a string of amino acids to a final, folded protein is still really really hard.
Well, if you don’t know, how are we supposed to know? I see high temperature and high pressure discoveries all the time on the net. Low temp stuff too. Like superconducting.
Stop making an ass out of you and me.
I am a high school chemistry teacher. Hardly a studied chemist.
Just because something new is discovered doesn’t mean it can’t be predicted by modelling.
It sounds like for protein folding, the multitude of forces involved makes predictions difficult.
New reactions, new compounds, new states, new properties are being reported in the scientific literature all the time. Buckyballs, e.g., were first discovered only 30 years ago - although they had been predicted earlier - and that launched a major new area in chemistry.
Models and theoretical arguments can do a lot, but they don’t supersede actual experimentation. Both are needed and both lead to unexpected results.
This might not be fully in the chemistry wheelhouse, but the metal whisker effect isn’t very well understood at the moment.
We know it happens with tin and some other metals (Zinc is another), and we know that adding a certain amount of lead inhibits it, but we don’t know why that is. With wider and wider adoption of the Reduction of Hazardous Substances initiative, which mandates the use of lead-free solder, we will need a lead-free solution to this problem very soon.
Well Linus Pauling got the Nobel for showing that all chemistry as we term it is governed by QED. John Pople and Walter Kohn got the Nobel for kick starting ab-initio computational chemistry, by getting the QED into a shape where you could compute it.
Trouble is - ab-initio computational chemistry is stupidly hungry of compute. It needs to model each electron’s interaction with every other electron, including all possible energy levels. Clearly some approximations are needed, even so, it is just plain awful. As a rough approximation, the amount of compute needed goes up as the fifth power of the number of atoms. This sorely limits the size of systems. Moore’s Law means you get a doubling in atoms possible about every 8 years. That isn’t exactly great.
So, as alluded to above, you need semi-empirical and looser approximations if you want to do computational work with big molecules. Computational work that models biological systems is still silly complex. You typically need to surround your reaction with water molecules, and every single one needs to be modelled, including all the interactions between them, as well as with what it is you are modelling.
The protein folding problem was one of the computational grand challenges back in the 90’s, and it won’t be getting better any time soon.
I gather you missed “Breaking Bad.” We expect our high school chemistry teachers to be capable of anything. Instead we get “Johnny’s gone forever, he’s gone forever more. What he thought was H20 was H2SO4.”
You can also turn easy questions into very difficult questions just by turning them around. It’s (relatively) easy to answer a question like “Is this substance a superconductor, and if so, at what temperature?”. It’s very, very difficult to answer a question like “What substance is superconducting at room temperature?”.
Especially within the field of organic chemistry, there are practically infinite watys atoms can be joined. Remember those tinker-toy-like stick-and-ball molecular models from high school and college? Nature acts like a toddler, sticking those suckers together into a bewildering variety of shapes and combinations. I had friends working in organic chem who gave fanciful names to these structures, based on what they looked like or structurally examined. Buckyballs and nanotubes are just some of the sexier versions of these, notable because they’re made up of carbons in an unexpected form. But if you allow other atoms in the mix, you can go wild with the combinatioons. And for each of these, there are a whole series of reactions (especially how to consciously fabricate them) and properties to explore.
And, mind you, this is besides the protein-folding described above, or polymer chemistry, or whatever.
There’s plenty to keep people busy for a long time.
12 grams of carbon contain about 6 x 10[sup]23[/sup] atoms. The fastest supercomputer in the world can perform about 3 x 10[sup]16[/sup] operations per second. That means to even enumerate every atom at 1 operation per atom would take about 200 days.
We’re pretty far from accurate simulation of macro scale phenomena. Everything we do uses hacks and shortcuts to optimize the running time down to something feasible.
I ran across a game called foldit awhile back. It spreads tries at protien folding among thousands of computer users. Does anyone know if it’s made any breakthroughs in how to model protein folding?
That game doesn’t make breakthroughs; it is a breakthrough. One solution to the problem of “this is a lot of work” is to get a lot of people to do your work for you. But it won’t make it any easier.
This is an aside, but I have had a lot of fun reading the blog, Things that I won’t work with. In it you find all sorts of crazy stuff that only chemists with a death wish attempt. Of particular interest is fluorine chemistry. (FOOF anyone? Perhaps some chlorine trifluoride then.) It appears there is an opening for someone who wants to investigate the sulfur chemistry of FOOF.
Predictions are one thing, but the practice often yields surprising discoveries.
I saw something interesting a few months ago.
Some chemists created a massive database of organic synthesis reactions, detailing starting materials, intermediates, steps, yields and catalysts. Then the computer sorted the data and started creating new links. By doing so, they were able to identify alternative pathways and shortcuts through from start to finish, and new ways to explore production for many reactions.
I also met someone who studied organo-fluorine chemistry. If you have a biologically active drug, it has a useful half-life in the body. Sometimes, this is too short to be really useful. Often, though, you can make it harder for the body to break down by adding things like fluorine and chlorine, without impacting (or sometimes enhancing) the useful features of the drug. But to take a complex organic molecule which is already hard to synthesise, and then try to do the same thing while sticking a halogen on to it (but not right in the biologically active bit) is some very difficult chemistry. And then do it again somewhere slightly different, just to see if it is any better.
We still have lots to learn in the chemistry world.
Sometimes, simpler models work as well or better (defining “best” as “gives the result which is closest to actual experimental values”), but you don’t know what will be the case until you run simulations on several methodologies for something for which you do have experimental values. Once you do that to figure out which methodology to use, you can start playing with the stuff for which you do not have experimental values. (Disclaimer: it’s what my papers were about).
One of the issues is the treatment of solvent. These are often not represented as individual molecules (which would raise the amount of atoms very quickly), but as a system constant. With apolar solvents this works pretty decently; with polar solvents such as water, it doesn’t work so well, and when you’re modeling biochemical processes (reactions, protein folding… there is a whole enormous field in nucleic acid modelization which could make protein folding look like baby legos) it often turns out that individual water molecules are a necessary agent. It’s not “just” the medium for the reaction, but neither is it a reagent in the traditional sense, it forms hydrogen bridges between different points of a protein, or between different molecules, thus helping get things in the necessary configuration. So… how many individual molecules do you actually need to paint, to get your results right? At least it will be as many as X-Rays say there are trapped inside the solid protein, but it can be more.
So how many amino acids does a protein have to have before its quaternary structure will choke a modern supercomputer? Would hemoglobin do the trick?
Hemo was already being modelled 20 years ago, in computers that were smaller than your cellphone. Modelling it’s folding, on the other hand, is what’s a lot trickier.