Combining chemistry with computers

[CrimpJiggler]

For years I had the ambition to get a degree in chemistry and possibly become a chemist because I like the idea of being able to manipulate chemicals but when I sit down to do some learning I find I always end up spending most of my time learning about computers and IT because I have a compulsion to learn that kinda stuff and find it way more enjoyable. I think my real passion is for computers and technology. I don't just think it, I know it. I'm already half way through this chemistry course though so I'm going to just finish it and get a BSc but since computers and technology is my real passion I need to find a way to implement it in chemistry. I have plenty of ideas and believe that in the future the majority of the work done by chemists in the lab will be done by machines so I may as well get involved in developing these machines and the software to control them but I'm wondering if I should get a 2nd degree in computer science so I'll be qualified (on paper at least, I know I'm more than capable of excelling in this regardless of whether I have a degree or not) to get into these kinda fields. Any advice?

 

[Mike H]

So, the obvious thing I thought right before I opened this thread - "Well, of course, there's computational chemistry and chemoinformatics for you to consider." However, it seems that your interests may not precisely overlap with that.

Now, I don't know precisely what you have in mind when you talk about the majority of the work will be done by machines, but I presume you mean above and beyond the ever-increasing amount of work that already is done by a wide variety of instruments and devices, including high-throughput technology. I mean, we already have autosamplers for NMR spectrometers, biosensors that can be programmed to run overnight and screen a whole truckload of different samples, and so on. Perhaps if you could elaborate a bit more, that would be helpful.

One of the goals that is out there is the idea of automatic structure verification - basically, how to write software that will be able to automatically determine the structure of a molecule via NMR spectra for small molecules. This would be a huge boon to the chemical/pharma industry, and let the people that they hire to do this work on the more difficult problems. This would be a case where you'd have to fold in more traditional computational chemistry/chemoinformatics, as you'd want to take advantage of the knowledge there, since it'd be essential to get such a project to work.

 

[Borek]

One thing to remember is that most of things done with computers in chemistry is based on on numerical methods and their applications. This is part of IT, but a specific part - and unless that's what you find interesting, it will be difficult to combine chemistry and computers. If it IS the part you like, you probably already know that what you will need is as strong background in math as possible, and a good understanding of quantum methods. Paradoxically, that sometimes mean chemistry background is less helpful in computational chemistry than math and physics.

 

[CrimpJiggler]

Now, I don't know precisely what you have in mind when you talk about the majority of the work will be done by machines, but I presume you mean above and beyond the ever-increasing amount of work that already is done by a wide variety of instruments and devices, including high-throughput technology. I mean, we already have autosamplers for NMR spectrometers, biosensors that can be programmed to run overnight and screen a whole truckload of different samples, and so on. Perhaps if you could elaborate a bit more, that would be helpful.

I'm not a chemist yet so I don't know what actually goes on in a real chemistry lab these days but what I have in mind is I think that the majority of tasks done by the average chemist can be done by machines. Done more effectively at that. For example titrations for analytical chemistry. This kinda thing could easily be automated. The chemist would only need to add the titrant to the burette and sample to the analysis chamber. The machine would use spectrophotometry or other methods to determine the end point. Another example is determination of solubilities. A machine could be built which would hold a range of common solvents and would have a series of chambers to which different samples can be added. Let's say for example you want to accurately determine the solubilities of a series of aniline salts. In the 1st chamber you add the free base. In the other chambers you add various salts. The machine will then take a small portion of each sample, and test its solubility in a particular solvent at a range of different temperatures. The machine will use spectrophotometry to determine the degree to which the compound is dissolved and will gradually raise the temperature of the solvent from just above its melting point (within reason, obviously it would be difficult for the machine to cool a solvent with MP of 15K etc.) to just below its boiling point and considering it will have measured the solubility continuously as the temperature rose, it will be able to construct a highly accurate solubility curve. If this kind of technology existed (or was at least widespread) then maybe we would have some decent publicly available solubility databases on the internet. The situation right now is ridiculous. In order to find the solubility of a compound you have to search through the CRC handbook which will only tell you things like "partially soluble in water at 25C, very soluble in ethanol at 25C" etc. which is nowhere near adequate information.

One of the goals that is out there is the idea of automatic structure verification - basically, how to write software that will be able to automatically determine the structure of a molecule via NMR spectra for small molecules. This would be a huge boon to the chemical/pharma industry, and let the people that they hire to do this work on the more difficult problems. This would be a case where you'd have to fold in more traditional computational chemistry/chemoinformatics, as you'd want to take advantage of the knowledge there, since it'd be essential to get such a project to work.

Yeah I think chemists make pretty good use of computers when it comes to spectrometry. Could be a lot better in my opinion though. The software we have for the spectrometers at my college isn't very good at all in my opinion. I'm not saying I could do a better job of making it but I could certainly expand upon it and improve it. Is this kinda stuff the domain of computational chemistry? I found out that they offer a computational chemistry MSc program at my university. I'm contemplating doing that once I've got my BSc.

 

[Borek]

For example titrations for analytical chemistry. This kinda thing could easily be automated.

http://www.google.com/search?q=automatic+titrator

Another example is determination of solubilities.

First, you have to collect samples of all substances, and that's already where it gets impractical. Technically it is not a problem, just the market is very limited.

 

[CrimpJiggler]

First, you have to collect samples of all substances, and that's already where it gets impractical. Technically it is not a problem, just the market is very limited.

What do you mean? Why would you need samples of "all" substances to use a machine like this? Do you mean if you were starting up a solubility database? In that case, yeah you would need samples of as many substances as possible to make the database as complete as possible but there's no reason why this database would have to be constructed by a single person/organisation.

 

[Borek]

Note that such a machine would give results in pretty short time. IMHO new substances are not prepared often enough to justify cost of such a machine in the average lab. It would make sense if it will be used often - that means either a huge lab, where many substances are being made and researched, or a specialized lab, which collects samples from various sources. Otherwise it will be underutilized.

I doubt nobody tried to design such a thing, I am not aware of any being offered - so I guess there are either a technical, or economical hurdles.

Edit: please note it is less problem of computers used, more a problem of mechanical/optical/electrochemical/whatever design.

 

[CrimpJiggler]

Yeah I see your point. What I have in mind though is using such a machine to devise a complete solubility database. A machine such as this would produce consistent, accurate results so this database would not just be complete, it would be trustworthy. The construction of such a database would make a very large contribution to the advancement of the field of chemistry though so I think that if this was to be implemented into some kind of project, this project would have the backing of chemists and people with a vested interest in chemistry from around the world. Like I said this does not need to be localised in a single lab, it could be labs all over the world making the contributions necessary to devise a relatively complete database of this sort. Whether they would or not, I believe government of various countries and other organisations should fund a project like this because the payoff would be fairly immense in my opinion. If universities all over the world were to buy this machine then this project could become feasible.

You're right though, the machine itself would not be of much use itself outside of contributing towards the construction of a complete solubility database. The construction of a database like this would contribute significantly towards the evolution of chemistry though so I think its well worth the investment required to build them. Imagine there was a website you could visit, in which you are 100% guaranteed to find detailed and accurate info on the solubility of whatever compound you have in mind in a wide range of common solvents. This website could give you nicely displayed solubility curves for single and even double solute systems. Thats where this database would really come in useful. Since its centralised, it would be feasible to add double solute systems to the database. For example if I want to know the solubility of KOH in a 1M solution of NaCl, right now I'd have to scour the internet and if I actually did manage to find the info, I wouldn't know if I could trust the source of the info or not. With a centralised system like this, people from all over the world could input their results so that the possibility of erroneous results in the database gradually decreases with time. The technology itself very simple to build, its just a matter of putting together various technologies that already exist and writing the appropriate software.

 

[Vanadium 50]

This sounds an awful lot like "a solution in search of a problem" (no pun intended).

I suspect that part of the problem is that chemists, at least the ones I know, don't spend a lot of their time measuring solubilities of inorganic salts. Or even needing to know that.

 

[CrimpJiggler]

Vanadium 50 that particular solution was motivated by my awareness of a problem. The problem is I can never find adequate solubility info on the internet. If there existed a complete, centralised, accurate database then I and many others wouldn't have this problem.

 

[Vanadium 50]

Yes, but this is like using a blunderbuss to kill a gnat.

First, the fact that you couldn't find the information in Google doesn't mean it doesn't exist on the internet. Much of the data that I need is positioned and indexed for the convenience of professional researchers, and not for the convenience of the general public.

Secondly, if you want something to catch on, you need to solve a problem that researchers are spending a lot of their time on. Solubilities of inorganic salts really don't fall into that category.

 

[Mike H]

First, I would be reluctant to characterize the current state of scientific software by what is used in undergraduate teaching labs. I've found that the actual research labs tend to have better toys, although certainly there is always room for improvement.

Writing software is not what I normally consider when people speak of "computational chemistry" - that phrase is generally understood, in my experience, to mean that you are investigating some chemical process by computational methods (e.g., dissecting the mechanism of a surface-catalyzed reaction). Certainly, new software can get written in the course of this work, but it's to solve the problem(s) at hand first and foremost.

Insofar as solubility testing, I believe that there are some automated solubility testing setups that have recently been utilized by the pharmaceutical industry to test various formulations, as that is one of the important criteria in bringing a compound to market.

Of course, have you seen the IUPAC-NIST Solubility Data Series?

http://srdata.nist.gov/solubility/index.aspx

I make no claims for its completeness, but I have found it helpful over the years to many questions. They also include more than just inorganic ions in their database, as well as ternary and quaternary systems.

I would also echo what Vanadium 50 mentioned - you need to find problems that are more than just trying to mine parameter space in the hope you'll find something. For example, there is still plenty of interest in the entire area of Hofmeister phenomena. If you needed to develop novel ways to do high-throughput solution screening of macromolecules with various ions in terms of solubility and other testing, that'd be perfectly understandable and very interesting to many people.