Slashdot Mirror

← Back to Stories (view on slashdot.org)

AI Plus a Chemistry Robot Finds All the Reactions That Will Work (arstechnica.com)

Posted by BeauHD on from the sit-back-relax-and-let-it-do-its-thing dept.

A team of researchers at Glasgow University have built a robot that uses machine learning to run and analyze its own chemical reaction. The system is able to figure out every reaction that's possible from a given set of starting materials. Ars Technica reports: Most of its parts are dispersed through a fume hood, which ensures safe ventilation of any products that somehow escape the system. The upper right is a collection of tanks containing starting materials and pumps that send them into one of six reaction chambers, which can be operated in parallel. The outcomes of these reactions can then be sent on for analysis. Pumps can feed samples into an IR spectrometer, a mass spectrometer, and a compact NMR machine -- the latter being the only bit of equipment that didn't fit in the fume hood. Collectively, these can create a fingerprint of the molecules that occupy a reaction chamber. By comparing this to the fingerprint of the starting materials, it's possible to determine whether a chemical reaction took place and infer some things about its products.

All of that is a substitute for a chemist's hands, but it doesn't replace the brains that evaluate potential reactions. That's where a machine-learning algorithm comes in. The system was given a set of 72 reactions with known products and used those to generate predictions of the outcomes of further reactions. From there, it started choosing reactions at random from the remaining list of options and determining whether they, too, produced products. By the time the algorithm had sampled 10 percent of the total possible reactions, it was able to predict the outcome of untested reactions with more than 80-percent accuracy. And, since the earlier reactions it tested were chosen at random, the system wasn't biased by human expectations of what reactions would or wouldn't work. The research has been published in the journal Nature.

39 comments

  1. But can it write its own research papers? by petes_PoV · · Score: 4, Interesting

    And equally important: can it be networked to similar machines (preferably made by other manufacturers and run by different labs) to set up its own peer reviews?

    And how soon before the drug cartels are buying up every machine that is produced to discover new substances?

    --
    politicians are like babies' nappies: they should both be changed regularly and for the same reasons
    1. Re:But can it write its own research papers? by Anonymous Coward · · Score: 0

      And equally important: can it be networked to similar machines (preferably made by other manufacturers and run by different labs) to set up its own peer reviews?

      And how soon before the drug cartels are buying up every machine that is produced to discover new substances?

      10 MONSANTO = TOXIC#0
      20 Goto 10

    2. Re:But can it write its own research papers? by Anonymous Coward · · Score: 0

      Well I work for a 3 letter agency and last week I seen some new forms that will need to be filled to get one of those so we should be safe from drug cartels.
      --
      Dwayne Johnson's Rampage As A Kaiju ("Weird Beast") Monster Movie

    3. Re:But can it write its own research papers? by Faluzeer · · Score: 4, Funny

      The more pertinent question, can it fake its own data for its research paper ;-?

    4. Re:But can it write its own research papers? by Anonymous Coward · · Score: 0

      This part is already solved

    5. Re: But can it write its own research papers? by Anonymous Coward · · Score: 0

      Hidden Figures taught us that computers don't author reports.

    6. Re:But can it write its own research papers? by Anonymous Coward · · Score: 1

      No, but it can find reactions for revolutionary new battery tech and simultaneously complain about how it'll never work on slashdot.

    7. Re: But can it write its own research papers? by LifesABeach · · Score: 1

      I was thinking of uses applied to grading plants.

  2. Kryptonite by Anonymous Coward · · Score: 0

    I hope it can figure out to make kryptonite!
    LOL :)
    --
    Dwayne Johnson's Rampage As A Kaiju ("Weird Beast") Monster Movie

  3. Automated industrial design by Anonymous Coward · · Score: 1

    I wonder if it's possbile to automatically design an industrial scale chemical process, an entire factory complex?
    Say I want to produce x amount of this or that product. What raw materials do we need? Where should we build the factory? What would it cost? What are the options?
    The system should of course take into account costs, externalities and regulations. Automatic procurement of commodity materials should be possible as well.

    1. Re:Automated industrial design by Richard+Kirk · · Score: 3, Interesting

      It would have to learn how to do the chemistry at factory scales to do that. This is possible but unlikely to be economic.

      If anything, the interesting work would be to go in the other direction. It is possible to make lab-on-a-chip units where the reagents are sealed in and each experiment is done in a small drop. This means you can repeat each experiment many times and see whether the results are repeatable. If your process is catalysed by some low concentration impurity, then that effect will vary with the number of molecules of that impurity, and you would expect a larger scatter in the yield.

    2. Re: Automated industrial design by Anonymous Coward · · Score: 0

      If you can give a score to the every result AI comes up with, you can train AI to do it.

  4. Interesting by wierd_w · · Score: 4, Insightful

    A carefully selected group of these working in parallel could theoretically parse the entire possible set of reactions, given sufficient time. (Yes, I know that with infinite molecular weight, there is an infinite number of possible compounds. However, only so many heavy molecules are interesting or useful, and of those, there will be certain classes that are more interesting than others. This approach would permit investigation of pathways without actually expending reagents, once its models are accurate enough. That means after a certain amount of training, a theoretical molecule of interest could be presented to the AI, and it could shit out the ideal synthesis pathway, and the next efficient arbitrary "n" pathways.)

    This is the kind of thing that is the beginning of universal replicators.

    1. Re:Interesting by SCVonSteroids · · Score: 0

      and it could shit out the ideal synthesis pathway, and the next efficient arbitrary "n" pathways.)

      Of all the places it could produce information, does it really need to come out of it's anus?

      --
      I tend to rant.
    2. Re:Interesting by tkotz · · Score: 1

      I think all compounds below a certain arbitrarily high molecular mass is a lot larger of a search space than you think. And if you want to try any possible experiment you really need one large machine more than a bunch of smaller machines however a matrix of smaller machine could fill in holes in the knowledge base for limited number of reactors. the number of possible reactions input combinations is 2^N where N is the number of known compounds, which will go up as the reactions uncover new compounds. It is estimated that we have already discovered 50 million different chemicals. So you are looking at a search space on the order of all possible 16-bit color 1920x1080 HD images, or brute forcing a 50 million bit crypto-key. and this is assuming all procedures are take equal amounts of all ingredients and mix them together at STP. You could greatly reduce the number of required input chemicals by only requiring each element to be represented once, but then you need to be able to conduct a lot more complicated procedures to synthesize intermediary chemicals.

      Also large molecules (like DNA, proteins and nanotubes) offer a lot of real world function and arbitrary complexity needed for more control over chemistry.

      This is a cool technology I just don't think it is realistically possible to cover "the entire possible set of reactions" the search space is just too big, and there will always be ways to take take what you've learned from the search to expand the search space.

    3. Re:Interesting by tlhIngan · · Score: 1

      A carefully selected group of these working in parallel could theoretically parse the entire possible set of reactions, given sufficient time. (Yes, I know that with infinite molecular weight, there is an infinite number of possible compounds. However, only so many heavy molecules are interesting or useful, and of those, there will be certain classes that are more interesting than others. This approach would permit investigation of pathways without actually expending reagents, once its models are accurate enough. That means after a certain amount of training, a theoretical molecule of interest could be presented to the AI, and it could shit out the ideal synthesis pathway, and the next efficient arbitrary "n" pathways.)

      This is the kind of thing that is the beginning of universal replicators.

      And the end to drug patents.

      You see, drugs aren't patented - you can patent the process to making the drug. Or more correctly, the process to make a something that will treat something. That's how drug companies often keep a drug patented - by finding a new use for it and keeping the basic process the same, so to treat the new condition you can't use a generic (even though they are identical).

      It's why generic drugs are so widely available - once the patent expires, the recipe is effectively open

      But with this robot, you know what the target is you want, and it can find an alternative way to make the same drug, bypassing the patent.

    4. Re: Interesting by Anonymous Coward · · Score: 0

      I do not think I have ever seen a statement so incorrect.

      Drugs certainly are patentable, generics are what happen when those patents expire and another company uses the same or different techniques to produce the same patented molecule.

      Full disclosure: I am a researching biophysicist who works with medicinal drug researchers.

  5. somebody explain me by sad_ · · Score: 1

    why would you use AI with a 80% success rate for this?
    isn't chemistry all maths? can't you just calculate the same reactions and have a 100% success rate?

    --
    On a long enough timeline, the survival rate for everyone drops to zero.
    1. Re:somebody explain me by davide+marney · · Score: 1

      And, how does pattern-matching on 72 known products not just give you all the
      products that are similar to the known set? How is that testing "all possible" combinations?

      --
      "We receive as friendly that which agrees with, we resist with dislike that which opposes us" - Faraday
    2. Re:somebody explain me by Anonymous Coward · · Score: 1

      Simulating reactivity accurately requires absolutely stupid amounts of compute. Also FORTRAN: https://en.wikipedia.org/wiki/List_of_quantum_chemistry_and_solid-state_physics_software

      Really "crude" analyses to predict absorption spectra of small biomolecules took everything we could throw at it in 2010, and we weren't exploring molecular interactions at all. The algorithms can be O(n^8) or worse.

    3. Re:somebody explain me by Anonymous Coward · · Score: 0

      Chemistry is quite complicated in practice, with a shitload of exceptions from the rules-of-thumb they teach in intro classes. That's why people actually do the experiments too, instead of just synthesizing things on paper and being happy.

      Chemistry classes usually teach you some intuitive rules-of-thumb that approximate the rigorous quantum chemistry quite well, and have you memorize common reaction mechanisms and their approximate yields. But if you wanted a 100% success rate calculator, you'd had to do a many-body quantum chemistry simulation, and that shit's expensive; it's not happening on a large scale without a quantum computer. If you can circumvent such a simulation using machine learning, you might save time and money in total, even if you waste some chemicals on the way.

    4. Re:somebody explain me by jouassou · · Score: 4, Informative

      First of all, most chemical synthesis reactions don't give one product with a 100% yield. Some molecules don't undergo the reaction, leaving some of the reactants behind. Some molecules have multiple sites where they can be chemically substituted, producing a statistical mixture of products. And some molecules have multiple isomeres that can be produced in the same reaction, but then behave differently in later reactions. So at each step of a chemical synthesis, you're polluting the system with more and more chemicals that might interact with each other and surprise you at later stages of the synthesis. The machine might perhaps be good at figuring out such interactions between biproducts of synthesis reactions even if a human just told it what the main product of each synthesis step would be.

    5. Re:somebody explain me by Anonymous Coward · · Score: 0

      I was going to write something, but then I realized the short article explains exactly this very well.

      Chemistry, like most things, is not all math because in many cases the math calculations would take thousands of years to run simulations that can be done with a short physical experiment. Or simplifying assumptions are made to make the mathematical calculations tractable, but lower the accuracy of the final answer.

    6. Re:somebody explain me by AHuxley · · Score: 1

      How many scientists and technicians can a nation keep working over decades?
      How many really great super computer labs can a nation spare from its nuclear weapons simulations?

      --
      Domestic spying is now "Benign Information Gathering"
    7. Re: somebody explain me by Anonymous Coward · · Score: 0

      Yes, you could. Good luck trying to explain that to an engineer, though. 'Hey! I can do the same thing with a calculator!', does not fit into their sci-fi delusions.

  6. Inputs by msauve · · Score: 4, Interesting

    Be careful what you feed it.

    --
    "National Security is the chief cause of national insecurity." - Celine's First Law
  7. Why Patents are Absurd by pubwvj · · Score: 2

    This is a prime example of why patents are absurd and should be discontinued.

    1. Re:Why Patents are Absurd by Anonymous Coward · · Score: 0

      Explain why.

    2. Re:Why Patents are Absurd by pubwvj · · Score: 1

      All too much work that results in patents is simply a matter of methodically working through combinations. I have invented a lot of things that way. Then I put them into manufacture and sell the products. That is the proper way to make money from your ideas. Patents ban other people from having similar ideas and using them. Patents are a bad idea and are being used to stifle innovation rather than the original purpose of promoting innovation.

  8. Let me guess.. by muhula · · Score: 2

    it was able to predict the outcome of untested reactions with more than 80-percent accuracy

    The skeptic in me wonders if 80% of the combinations had no reaction

  9. And when it's done ... by PPH · · Score: 2

    ... does it pour all the products down the lab sink?

    --
    Have gnu, will travel.
    1. Re:And when it's done ... by s_p_oneil · · Score: 1

      It uses an "Undo" feature to reverse the reactions until it gets back to its base components.

  10. Complete garbage by Anonymous Coward · · Score: 0

    This is complete garbage. Finding chemical reactions is a search through a high dimensional space. AI algorithms do do such a search. Poorly. The kind of search that is reported involves finding the trivial. Finding something truly new requires a chemist's insight, just like any other creative activity. For example, what catalysts are tried? What are the limits to the ambient pressure?

  11. kind of interesting... by avandesande · · Score: 1

    But it is backwards chemistry. Nobody in industry has a bunch of stuff on the shelf and says 'what can I make with this'. There is a chemical they need to make and they ask themselves 'how can I make this'.

    --
    love is just extroverted narcissism
    1. Re:kind of interesting... by bugs2squash · · Score: 1

      But so many things have been stumbled upon - like Saccharine. The really fancy version of AI would suggest possible uses for whatever it creates.

      --
      Nullius in verba
  12. Because chemistry is HARD by Immerman · · Score: 1

    >isn't chemistry all maths? can't you just calculate the same reactions and have a 100% success rate?

    Because chemistry is applied quantum mechanics, and the math is HARD. I think they finally managed to simulate the chemical properties of a hydrogen molecule, but last I heard that was the most complicated molecule we've been able to simulate accurately - and it's literally the simplest molecule that exists, by a wide margin.

    --
    --- Most topics have many sides worth arguing, allow me to take one opposite you.
  13. You know what they got from those? by Anonymous Coward · · Score: 0

    You know what they got from those?

    Mundane Potions

  14. Point that AI ... by PPH · · Score: 1

    ... at this guy's blog as a training data set. And then watch it reply to a query.

    --
    Have gnu, will travel.