• N&PD Moderators: Skorpio | someguyontheinternet

Pharmacology An AI chatbot told me how to make tianeptine a more potent opioid

This thread contains discussion about a Pharmacology-related topic

negrogesic

Moderator: BDD, OD
Staff member
Joined
Jul 21, 2002
Messages
12,556
Location
Negronesia
I was fooling around with an AI chatbot and asked them the following question, which they answered quite nicely to my surprise:



It also answered a host of other questions, far quicker than I could of looked up. For instance ki values and ic50's of various compounds etc.

Here's another example:



The chatbot is free to use. The url is visible in the second screenshot. I'd rather not type the link because the AI was quick to catch on. The second time i asked the question about modifying tianeptine it refused to answer and told me this:



I mean there is nothing mystical happening here, one could look all this stuff up, but it definitely seems like quite a high powered search engine.
 
I've never used it, but people were prompting it to give two replies, one using chatgpt guidelines, and then one as DAN or D.A.N. (Do Anything Now) and Dan gives non censored replies
 
While not wrong, these are very general answers.

Does anyone know if it's tianeptine itself or metabolites that have opioid activity? That long-chain carboxylic acid being cleaved to leave a primary amine would see the scaffold overlay dinortilidine.
 
Try using a previous prompt, it really helps getting more precise answers. Something like "act like a senior medicinal chemist who is an expert on the SAR of opioids, I will provide you with X information and you will give me your top suggestions regarding XXX..."
 
Did you know that nobody has yet been able to come up with a set of rules that will tell you if a particular drug will be an opioid? With monoamine modulators (from MDMA to Prozac) it's possible to build rules... but every time we THINK we have a rule for opioids, we find one exception..

That's why opioids interest me. Every other class of psychoactive drug it essentially possible to calculate.. I mean, how DARE tianeptine be an opioid! Every time I learn 3 things, 2 of them are that I do NOT know something. Essentially I feel less sure all the time.
 
Did you know that nobody has yet been able to come up with a set of rules that will tell you if a particular drug will be an opioid? With monoamine modulators (from MDMA to Prozac) it's possible to build rules... but every time we THINK we have a rule for opioids, we find one exception..

That's why opioids interest me. Every other class of psychoactive drug it essentially possible to calculate.. I mean, how DARE tianeptine be an opioid! Every time I learn 3 things, 2 of them are that I do NOT know something. Essentially I feel less sure all the time.

Well there's the "morphine rule":

(1) a benzene ring (2) attached to a quaternary carbon connected by (3) a two-carbon spacer to (4) a tertiary amine.

But this certainly doesn't define all opioids
 
I asked a comparative SAR question and it gave me a nice answer:



This thing is pretty cool, I've never interacted with something that I could ask questions like this to. Perhaps I'll get the chip implant of this thing when it comes out in the year 2041 🤔
 
I think it's missed the fact that the larger groups is a sulfonamide - an N-methyl sulfonamide to b be exact. So again, I do not think wrong (look at methadone analogues that have a diphenyl moiety). It's avoiding the most basic opioid that was made due to logical design (methadone).

I suppose I could be wrong (quite possible) or the program still does not recognise key moieties. I was able to provide an overlay... they didn't/

So maybe feed that in and see what it says.

If they can access articles, journals and patents then I can USE this.

Many thanks, BTW. I know I can be standoffish - I'm working on it. It's ME not being good with people.
 
Damn man, I don't understand much but that is frigging awesome. If this is one of the first versions, imagine what it would be capable in 10 years. Probably you'll just say "R2, make me 10g of diamorphine" and in 12 hrs u got ur drugs.
 
Yeah - well I found U-47700 in a patent and that proved to be good. From that I designed U-93951 which is x22 morphine, as cheap and as simple to make as U-47700 and has a TI better than it's parent.

But I suspect that after fentanyl analogs, the next high-power opioid will be a thienorphine analog thus:


On the left is China's new treatment for opioid dependence (thienorphine). It's about x300M m potency with a duration of 2-3 days.

On the right is the FULL AGONST which is about x8000 morphine with a duration of about 12 hours.

OK so it is NOT cheap to make. But if ⅛ of a mg is as potent as a hit of heroin (and as euphoric) but a user only needs 2 hits a day to stay well. It does not take a lot of imagination to work out the value. Organized crime can have people growing the Chinese poppy (or other oripavine yieldine strain) and produce the stuff. Because the labs do not need to be big or need a lot of staff - 2 or 3 good chemists getting $100,000/year can make it.

The same argument was made with carfentanil. Well, people could make it BUT the idiots who paid for it tried to just make a powder. You totally have to make a uniform tablet form. Ideally like Temgesic i.e. sublingual because that means people can cook and shoot.

So in truth, the criminals are not trying TOO hard. If 200 million of these tablets (i.e. 1Kg made into pills) turned up in the US one day, the 'drug war' would just get more intense. The DEA would announce 30 years for people selling this particular compound and get a HUGE input fr om central government. Of course it would still become the de facto standard street opioid and people would die...

But that is 'the game'. Honestly, I've known a lot of people who do this kind of thing and they REALLY do think if it as a game. Meanwhile end users die.

Sorry to ramble on but I recall in the late 1980s people were saying that carfentanil would be the 'ultimate' opioid and defeat the DEA but the idiots who cut and sold it had NO IDEA of how potent it was. If instead pills had turned up, deaths would be much lower and the DEA would have lost. In fact, I have wondered if the DEA introduced carfentanil in a clumsy manner to cause death and make users mistrust the stuff.

I've had the DEA feel my collar and was amazed at their lack of finesse.
 
Chat gpt will answer completely fictional things that look totally right though, you have to validate what it gives you.

I've asked it biochemistry questions, and its answers are often generally feasible but it will also name amino acids not in the right spots on proteins and I have had it suggest fictional journal articles, where the DOIs correspond to a totally different article.
 
Chat gpt will answer completely fictional things that look totally right though, you have to validate what it gives you.

I've asked it biochemistry questions, and its answers are often generally feasible but it will also name amino acids not in the right spots on proteins and I have had it suggest fictional journal articles, where the DOIs correspond to a totally different article.
It does this with math proofs as well.

It's just an unavoidable consequence of how it works, i.e. choosing the next word based on the probability of it appearing given the prior word choices.
 
I've asked it biochemistry questions, and its answers are often generally feasible but it will also name amino acids not in the right spots on proteins and I have had it suggest fictional journal articles, where the DOIs correspond to a totally different article.
Yeah that's one of the current limitations of chatgpt as a language model which doesn't have access to the internet.
It will definitely come up with fake articles so it can't be used for citations or things like literature reviews. The Bing version might be able to do it eventually though.

ChatGPT is still super useful if one knows how to use it (specific prompts are pretty powerful as well)
 
I have had it suggest fictional journal articles, where the DOIs correspond to a totally different article.

Yeah this was very frustrating, it was happening to me. I started asking it to provide citations along with my questions and it would sometimes end up extracting a subheading that is located within some related journal article and use that as a title for the reference, along with some funky DOI. It was frustrating since I could find no such article. Now I sort of see what it's doing, and I can eventually find the reference, but it's alot of digging involved.

I don't put too much faith in its answers, but it is definitely interesting to play with.
 
If they can access articles, journals and patents then I can USE this.

It does scan journals, not sure about patents but I bet it's possible that it does.

It is worth playing around with as it can be quite a time saver, when its accurate.
 
Yeah - well I found U-47700 in a patent and that proved to be good. From that I designed U-93951 which is x22 morphine, as cheap and as simple to make as U-47700 and has a TI better than it's parent.

But I suspect that after fentanyl analogs, the next high-power opioid will be a thienorphine analog thus:


On the left is China's new treatment for opioid dependence (thienorphine). It's about x300M m potency with a duration of 2-3 days.

On the right is the FULL AGONST which is about x8000 morphine with a duration of about 12 hours.

But wouldn't it also be fairly non-selective for the MOR like the related compounds, thus not particularly euphoric due to too much KOR activation? I imagine that is the issue with dihydroetorphine as well.

Then again, hyperselectivity for the MOR isn't a good thing either, like fentanyl 🤔

I wonder what the chatbot thinks the optimal ratio of MOR:DOR:KOR activation is in terms of producing a reinforcing effect. I don't think that has been studied much though 🤔
 
But wouldn't it also be fairly non-selective for the MOR like the related compounds, thus not particularly euphoric due to too much KOR activation? I imagine that is the issue with dihydroetorphine as well.

Then again, hyperselectivity for the MOR isn't a good thing either, like fentanyl 🤔

I wonder what the chatbot thinks the optimal ratio of MOR:DOR:KOR activation is in terms of producing a reinforcing effect. I don't think that has been studied much though 🤔

Well I know dihydretorphine is (or at least was) abused in China. It wasn't considered as good as H and only lasted 2-3 hours... But swapping that 7-alkyl with an ethylaryl is the key to the increased duration so my analogue would last longer. I would expect it to be subjectively much like dihydroetorphine.
 
Top