New Deep-Learning Software Knows How To Make Desired Organic Molecules (nature.com) 46
dryriver shares a report from Nature about a neural network-based, deep-learning software that is as good as trained chemists in figuring out what reagents and reactions may lead to the successful creation of a desired organic molecule: Chemists have a new lab assistant: artificial intelligence. Researchers have developed a "deep learning" computer program that produces blueprints for the sequences of reactions needed to create small organic molecules, such as drug compounds. The pathways that the tool suggests look just as good on paper as those devised by human chemists. The tool, described in Nature on March 28, is not the first software to wield artificial intelligence (AI) instead of human skill and intuition. Yet chemists hail the development as a milestone, saying that it could speed up the process of drug discovery and make organic chemistry more efficient. Chemists have conventionally scoured lists of reactions recorded by others, and drawn on their own intuition to work out a step-by-step pathway to make a particular compound. They usually work backwards, starting with the molecule they want to create and then analyzing which readily available reagents and sequences of reactions could be used to synthesize it -- a process known as retrosynthesis, which can take hours or even days of planning. The new AI tool, developed by Marwin Segler, an organic chemist and artificial-intelligence researcher at the University of Munster in Germany, and his colleagues, uses deep-learning neural networks to imbibe essentially all known single-step organic-chemistry reactions -- about 12.4 million of them. This enables it to predict the chemical reactions that can be used in any single step. The tool repeatedly applies these neural networks in planning a multi-step synthesis, deconstructing the desired molecule until it ends up with the available starting reagents.
Comment removed (Score:3)
Re: (Score:2)
We could get some crossover action too, toss in some prilosec and add flaming habanero cheese flavor.
Crystal blue persuasion (Score:5, Funny)
Re:Crystal blue persuasion (Score:4, Insightful)
Yeah, finding syntheses of controlled substances from uncontrolled precursors is an obvious use of this tech.
Re: (Score:2)
Not just overpriced gorge-level pharmaceuticals; the reason "Dex" is called "Dex" is because it's the right-handed (dextro) version of amphetamine. Active methamphetamine is dextro too.
Re: (Score:1)
"Will it help make crystal meth?"
Sure. And explosives for bombs, lots of them.
Re: (Score:2)
Re: and so it begins (Score:2)
How do you preserve free speech and the spirit of the Constitution in a time when things can be spoken into existence?
And now! (Score:2, Funny)
Ed: Allow me to introduce the famous seer, sage, and soothsayer, Ladies and gentlemen, Carnac the Magnificent [youtube.com]
Ed: I hold in my hands the envelopes. A child of four can see that these envelopes are hermetically sealed. They have been kept in a mayonnaise jar on Funk and Wagner's porch. Nobody(!) knows the contents of these envelopes, but you in your mystical way will divine the answers. Here is the first envelope. . .
Carnac the Magnificent: (Holds envelop to head) . . . As a kite..
Ed: As a kite
Carnac
So, how about all those biotech jobs (Score:2)
Re: (Score:2, Informative)
They were replaced a long time ago. Most biotech labs are now a director, plus a few technicians to supervise the automatic sequencing machines. It used to take a PhD graduate three years of research to figure out what genes interacted with which proteins. Now that's done automatically. The cost of sequencing a single human genome has dropped from $100 million to under $1000 in 15 years. That's faster than Moore's law.
https://www.genome.gov/sequencingcostsdata/
So they moved on from genomics to proteonomics,
Significant and usefule, but ... (Score:3)
No doubt this is potentially a highly significant development, and an early example of a powerful tool that shows the way to the future. I expect that this sort of technology will prove useful in developing many desirable chemicals for many purposes. But, one of the things I wonder about is the potential for reduced understanding and insight among the people using it, and where it might lead. Mathematics is already confronted by machine generated results that are beyond the ability of humans to check. And I remember reading of results that seemed to be correct, but the method that they were arrived at was impenetrable. Trust the machine(s)? How far? Is this another area where AI might prove dangerous to humanity?
Computer generated math proof is too large for humans to check [phys.org]
Chess computers are now pretty much able to beat any human. Amazingly now computers playing Go seem to be heading in the same direction. Brute strength and clever algorithms combine to search possibilities far beyond what a human can. Someday will AI search out a subtle "final solution" for humanity that will take 10 generations to come to fruition? Checkmate?
How can we safeguard our future from subtle, malevolent AI?
Re:Significant and usefule, but ... (Score:5, Interesting)
But, one of the things I wonder about is the potential for reduced understanding and insight among the people using it, and where it might lead.
To give you an idea of the present state of chemistry, we only recently measured the energy of a transition state [phys.org], imaged atoms and molecules, or directly observed hydrogen bonds [phys.org]. New insights into the behavior of water is common reading. As for syntheses, the reaction mechanisms drawn are at best guesses and many times syntheses reasonable in theory are found not to work in practice. Basically, we chemists do not have much fundamental understanding so much as a practical intuition for how chemical systems behave. But improved understanding and classification often go hand-in-hand in science, so I think it likely that the output of these algorithms will actually improve our human-level interpretations.
Probably not "Deep Learning" (Score:1)
This sounds like a problem involving scanning databases of known reactions, and doing detailed modeling. There might be an artificial neural network in there somewhere but I doubt it is the key.
It reflects the sad state of science journalism that anything vaguely intelligent becomes "Deep Learning", or whatever the current buzz phrase is.
Re: (Score:2)
How can we safeguard our future from subtle, malevolent AI?
Learn how to make simple tools and how to use them.
Re: (Score:2)
Through history, mathematicians have invented new ways to abstract and notate ever more complex mathematical problems. What began with simple addition, then multiplication - exponentiation - summation - calculus - differential equations - .. (I stop here because that's about my limit of understanding... I'm not a mathematician, and the last category I mentioned I'm not even very good at... but I know there are some further abstractions 'out there'.)
Computers/various processors are very good in doing a lot o
Making molecules is not the bottleneck (Score:2)
... I imagine, without knowing much about biotech (or RTFA). The bottleneck is the trial and experimentation, which takes a long time. You already have to be very discerning in deciding which synthesized compounds you want to try, what good does it do to you to be able to computer-generate more compounds?
Re:Making molecules is not the bottleneck (Score:5, Insightful)
what good does it do to you to be able to computer-generate more compounds?
As a chemist, I can tell you that there are different bottlenecks at different stages of the development. Early on in the process, when you are working a lab bench scale, and you synthesize in the gram range, you can probably work around some issues. Some key processes may be an issue and you have problems synthesizing the right compounds. Also, there's the problem of how many sequential steps you need to make. If your compound is simple and you need to mix only two different compounds, then a 90% yield is fine. But if each step gives you 90% yield and you have 10 steps in a row you're down to 35%. Now imagine how many steps a compound like Taxol [wikipedia.org] would take if you started from scratch...
Finally, there's the large scale process, the one that allows you to produce at industrial scale, think kg or tons, the molecules that survived the trial process. Here, you have a bunch of other considerations. Health hazards of the precursors (it's much simpler to work with stuff that has no environmental/hazards hazards, than on a fully sealed line), safety hazards (if the reagent burns in contact with air or water, it's probably a bad option), cost, yield, etc. At this step the reactions are typically redesigned, because what works well at the bench scale, doesn't work at industrial level. So, if you can do this step with a computer and it gives you a better option, it's a benefit, even if you're applying it only to the molecules you already know.
v2.0 will assess the organic molecules' traits (Score:1)
Interesting analysis (Score:2)
I recently went through all of Derek Lowe's "Things I Won't Work With" [sciencemag.org] columns (highly recommended for anyone with a sense of humor and an instinct for self-preservation), and in the aftermath spent some time reading some of his other articles. One in particular discusses the possibility of an automated chemist [sciencemag.org], performing reactions given a recipe. Today's article [sciencemag.org] discusses this latest paper, which focuses on generating those recipes, and compares it to another AI approach [sciencemag.org] previously covered.
Notably, Lowe f