15 ms·
With a nudge from AI, ketamine emerges as a potential rare disease treatment
- loceng 5y ago"The tool, called mediKanren, scanned millions of biomedical abstracts hunting for relationships between existing compounds and the gene involved in the disease." Is this AI or someone put the effort into compiling a database and added automatic search function?
- tartoran 5y agoReminds me of minikanren and I think this may be related
- dylan604 5y agoIsn't that all AI is in its current state? Just like computer chess. It knows the moves, and can calculate possibles and iterate through them way faster than a human. It's not like it is thinking for itself saying, what happens if I try this. The devs of the AI point it in a direction, and it sorts through things much more deeply and faster than humans. All of that is great when we get useable information or even when it exhausts all options to no positive result. It's just another tool to find out answers. The term AI is used waaaay too broadly.
- edejong 5y agoThat was the state of AI in the 90's. AlphaGo Zero and AlphaZero have shown that it does not need to search exhaustively when it can search intelligently using reinforcement learning. The 'direction' is something the system learns and is not given by the devs.
- cratermoon 5y agoTake out 'exhaustively', though, and everything parent comment said is correct. AI is just (big data + statistical inference) implemented in massively parallel ways using linear algebra.
- cortezdecoy 5y ago"just" is always a weird modifier in these kinds of conversations. Most things can be reduced to "just" some basic operation(s). What is amazing is the complexity that emerges from those simple operations at scale.
- cratermoon 5y agoGranted that emergent behaviors are important. The reason for injecting "just" is to counter the AI hype. Imagine if the headline read "With a nudge from linear algebra, ketamine emerges as a potential rare disease treatment". Dropping the word AI in there, in addition to boosting the articles SEO, will suggest to those with a casual interest that his must be a big deal, because "I've heard so much about how AI is solving <insert big problem here>"
- edejong 5y agoWell, if you dive into data science and ML, you'll see it is much more than just a bit of linear algebra. Some topics (and as a data engineer, I am not an expert): - Distributed computing - Feature selection - Budgeting / accounting of experimental design (these TPU clusters are not cheap) - ML architecture - Involvement of domain experts (multidisciplinary teams) - Storage The whole list is much longer, but just some topics to think about. For me the term "AI", although overly broad, come to mean the engineering and organisation needed to pull these kind of projects.
- doit4thebitties 5y agoI worked at a biomedical informatics shop that did this very thing by AI/ML 10 years ago. Essentially: automated metanalysis.
- rscho 5y agoI don't think that's automated meta analysis. It's meta research, which is not the same thing. Take a look at the MediKanren papers. This is logic programming and it's not based on statistics, but first order logic. In fact, it's a kind of expert system. But in the 80's, expert systems didn't have access to that: https://skr3.nlm.nih.gov/SemMedDB/ https://skr3.nlm.nih.gov/SemMedDB/
- rscho 5y agoThis is logic programming. So AI in the symbolic sense. And much better suited to medical 'AI' in the current state of medical information retrieval capability than most of the statistical crap that gets all the hype nowadays.
- pizza 5y ago> Proof-of-concept for reasoning over medical knowledge graphs, using miniKanren + heuristics + indexing. https://github.com/webyrd/mediKanren https://github.com/webyrd/mediKanren
- civilized 5y agoGitHub says it's mostly written in Scheme and Racket! Would love to hear the story behind that... is this just one researcher's idiosyncratic choice, or are these languages providing some kind of material advantage in this application? (I'm aware that Lisps have a long history in AI but I understood that to be kind of in the past at this point...)
- rscho 5y agoScheme is the ideal language for something like MiniKanren, because of its 'almost-purely-functional' nature and its powerful macros. MiniKanren is all lambdas at its core with clever macros sprinkled on top. And also, Dan Friedman, Matt Might and Will Byrd are VERY clever people. So much for people complaning nothing useful ever comes out of Racket!
- mattmight 5y agoI'm the researcher mentioned in the article. We used Racket in part because before we transitioned into academic medicine, Will Byrd and I were in academic computer science with a focus on programming languages. Will was building up miniKanren to do lots of cool stuff in program synthesis and program analysis. When we jumped over, it felt natural to bring the power of relational reasoning to knowledge graphs over biomedical knowledge, and so mediKanren was born.
- rscho 5y agoIm a big fan of your work, Mr Might! I'm just an MD with a hobby interest in logic programming, but FWIW I think that you guys have really captured how to use current medical information, a thing that the statistical approach still cannot achieve. Congratulations to you and your team!!
- cratermoon 5y agoUh oh, here come the ketamine hucksters.
- tomato-sauce 5y agoIt feels like in the last few years I have been constantly hearing about new research showing “fun” drugs as cures to a wide array of medical conditions. There’s marijuana for chronic pain, mushrooms for depression, mdma for ptsd, and now ketamine. They are often talked about as if they have little or no harmful side effects. I’m glad we are moving away from the incredibly destructive drug policies that have been in place for so long but I fear the pendulum may be swinging too far the other way. The opioid epidemic was largely caused by the idea of them being non addictive being heavy pushed by the pharmaceuticals industry. Are we going to see similar harms from these drugs a few years after they gain more popularity? So far the evidence doesn’t show anything near this level of harm. I’m just skeptical of anything that seems too good to be true.
- nickthemagicman 5y agoWhat there a really are no or very few side effects?
- shadowgovt 5y ago(interpreting as "What if there really are no or very few side effects?") That would be great, but it would be something we learn on the other side of controlled trials.
- IBCNU 5y agoThe idea of the pendulum swinging too far is a huge over-reaction. The pendulum has 40 years of swinging back to a "normal" state if you include the overprescribed industry of prescribing mood altering drugs to suburbanites, adderall, etc. Yes, the boomer and gen-x generations were raised in a lie: their doctors told them they could take drugs without side effects. However, this is now well known.
- narag 5y agoKetamine seems to be a commonly used drug in emergencies because it's a safer anesthetic than propofol when you don't know specifics about the patient and can't determine dosage.
- cupcake-unicorn 5y agoI recently got my full genome sequenced via a website called nebula.org. I don't have any backgrounds in genetics but I was playing around with the raw data with some github apps. Thanks so much for this - I may be able to link this app in to my custom data. That would be so awesome if we could hook this up to personal genetics results. Even people who aren't sick do have mutations and it would be so cool seeing AI running some kind of search on their personal genome and suggesting a drug for that! Here's the link to the github app they mentioned: https://github.com/webyrd/mediKanren https://github.com/webyrd/mediKanren
- Vecr 5y agoA drug for what? I don't see how running an AI on your genome to find drugs for a non-existent genetic disease you don't have because the AI just hallucinated it.
- JPLeRouzic 5y agoSome 15 years ago someone (sorry I forgot who) in Waterloo university published a paper about how to use a search engine to discover hidden relations. It was something like this: 1. Search for the key word "A" 2. Find new keywords (b,c,..) that are associated with "A" (I did this with lingpipe [0] 3. Search each the new keywords (b,c, ...) 4. Find for each of them their associated keywords (finding (s, t, ...) 5. Repeat 3. 6. repeat 4. finding (z, x, ...) 7. If in (z, x, ...) one of them is associated with "A" then you discovered an hidden relation between "A" and one of the members of (s, t, ...) (hidden if it did not appear in the initial search). [0] http://www.alias-i.com/lingpipe/ http://www.alias-i.com/lingpipe/
- rscho 5y agoI am surprised people here seem mostly interested in discussing drugs when the object of this article is the materialization of the wet dreams of all the ML fanboys out there. Reasoning over medical information that actually works! Isn't that remarkable?! Of course, it's symbolic and not statistical 'AI' so it doesn't tickle the deep learning hipsters announcing AGI as just around the corner.
- rscho 5y agoThe video that explains it all for techies: https://youtu.be/d-Klzumjulo https://youtu.be/d-Klzumjulo
- DoreenMichele 5y ago"In these patients, that gene is sort of running at 50% capacity: One copy’s broken, the other copy’s working,” explained Might. “So the tool, when it sees that, says, ‘OK, well, maybe I can make the functioning copy work twice as hard to compensate.’” Woo! Biohacking at its best. I do something sort of similar by eating hot peppers regularly. It opens up a different chanel than my defective chanel and helps me keep my chronic edema down to a dull roar, among other things. To prepare the kids and their parents, the Seaver center sent stickers to practice applying to the kids’ chests, so they could get comfortable with the sensory experience of ECG readings. I like these people.
- phist_mcgee 5y agoOn reading this article, it must be so hard for parents to see a way forward for treating their child's condition, but be roadblocked by either funding, or study participation criteria. If I observed my child have such a profound change as Mateo did, I think I might be incentivised to continue seeking these drugs for their treatment. I know that it would be absolutely a criminal act to purchase ketamine illegally and administer it to my child, or to find a shady doctor who prescribe it off-label. I wonder how often these scenarios come up? And how often parents go out of their way to seek unapproved medicines (much like the AIDS Buyer's Clubs of the 90s).
- sinuhe69 5y agoYes, other fathers did https://youtu.be/FSdjvEnyRzk https://youtu.be/FSdjvEnyRzk
- pneumonic 5y ago“I advise strongly against trying to pursue ketamine, commercially or clinically, because of safety concerns and the risk of losing our opportunity to study it rigorously,” said Kolevzon [the clinical director] Research to gain quality evidence is good, but sometimes other considerations outweigh that. Visible improvement in the sufferers of a rare disease may be such a case.