ChuangTzu
MAPS.org
Mathematical hallucinogens (or, Untested hallucinogens of possible high potency)
Thanks to Fastandbulbous, I'm reading over a fairly recent article (Shulze-Alexandru et al., Quant. Struct.-Act. Relat., 18 (1999)) on a model for quantitative structure-activity relationships of hallucinogens. The field (QSAR for short) proposes to predict the efficacy of unknown compounds by creating mathematical models based on structures of known activity.
The basic procedure for creating such models is to list any and all factors which may contribute in some way to the potency, or lack thereof, of a class of drugs, catalog said properties and potencies, and run them through a computer looking for correlations. Models differ in the parameters they choose to use as predictors. Almost any property you can think of has already been applied to the phenethylamines (thanks to Shulgin, QSAR studies with the phenethylamines are among the most solid in the field since there is so much input data to work with) from HOMO-LUMO energies to positions of maxima in the UV spectrum of the molecule. As computer time becomes cheaper and algorithms better, more and more input parameters are being used in combination giving the resulting models higher predictive accuracy. Sometimes a simple least squares method is used, sometimes a more sophisticated non-linear regression. Even neural nets have been applied to the problem.
This particular model attempts to incorporate quasi-atomistic receptor modeling the goal of which is to pay more attention to the actual receptor topology . 23 compounds (rather few) were used as training data and consisted of phenethylamines, tryptamines, and LSD. Seven compounds with known activity were used to test the model, with the worst prediction off by a factor of 2.7 in potency. Finally, the model was used to predict the potencies of 53 untested (at the time) molecules (including some hemi-fly and dragonfly-esque characters).
I don't have time at the moment to give a full summary of the results. I'll update this thread with more info as I have time, or other people with access to this article can feel free to chip in at will.
The most interesting prediction the authors made is for this compound:
,
estimating a Ki value of 3.2nm (for comparison, DOI has a Ki of 6nM---lower values indicating more potency). This prediction is interesting because of its structure. As far as I know, no phenethylamine has been synthesized and tried in man with any kind of amide functionality. Furthermore, even with a possible error of about 3-fold (the study gives an error of plus-minus 1.8, but let's be realistic), this compound is still well within the active range. Hell, even if it's off by a factor of 10, we still have a somewhat active compound. It's also interesting because, given a bit of DOI and some chemistry background, it would be a cinch to synth.
Of course there are a million and a half reasons why it wouldn't be active and anyone's uncle could think of at least 10. The most damning is that the compound looks somewhat like a hybrid between the ergoline and phenethylamine skeletons and only one (and one of the most potent) ergoline was used in the training data. But there's really only one way to find out
Thanks to Fastandbulbous, I'm reading over a fairly recent article (Shulze-Alexandru et al., Quant. Struct.-Act. Relat., 18 (1999)) on a model for quantitative structure-activity relationships of hallucinogens. The field (QSAR for short) proposes to predict the efficacy of unknown compounds by creating mathematical models based on structures of known activity.
The basic procedure for creating such models is to list any and all factors which may contribute in some way to the potency, or lack thereof, of a class of drugs, catalog said properties and potencies, and run them through a computer looking for correlations. Models differ in the parameters they choose to use as predictors. Almost any property you can think of has already been applied to the phenethylamines (thanks to Shulgin, QSAR studies with the phenethylamines are among the most solid in the field since there is so much input data to work with) from HOMO-LUMO energies to positions of maxima in the UV spectrum of the molecule. As computer time becomes cheaper and algorithms better, more and more input parameters are being used in combination giving the resulting models higher predictive accuracy. Sometimes a simple least squares method is used, sometimes a more sophisticated non-linear regression. Even neural nets have been applied to the problem.
This particular model attempts to incorporate quasi-atomistic receptor modeling the goal of which is to pay more attention to the actual receptor topology . 23 compounds (rather few) were used as training data and consisted of phenethylamines, tryptamines, and LSD. Seven compounds with known activity were used to test the model, with the worst prediction off by a factor of 2.7 in potency. Finally, the model was used to predict the potencies of 53 untested (at the time) molecules (including some hemi-fly and dragonfly-esque characters).
I don't have time at the moment to give a full summary of the results. I'll update this thread with more info as I have time, or other people with access to this article can feel free to chip in at will.
The most interesting prediction the authors made is for this compound:
estimating a Ki value of 3.2nm (for comparison, DOI has a Ki of 6nM---lower values indicating more potency). This prediction is interesting because of its structure. As far as I know, no phenethylamine has been synthesized and tried in man with any kind of amide functionality. Furthermore, even with a possible error of about 3-fold (the study gives an error of plus-minus 1.8, but let's be realistic), this compound is still well within the active range. Hell, even if it's off by a factor of 10, we still have a somewhat active compound. It's also interesting because, given a bit of DOI and some chemistry background, it would be a cinch to synth.
Of course there are a million and a half reasons why it wouldn't be active and anyone's uncle could think of at least 10. The most damning is that the compound looks somewhat like a hybrid between the ergoline and phenethylamine skeletons and only one (and one of the most potent) ergoline was used in the training data. But there's really only one way to find out
Last edited:
)
![:\](https://bluelight.org/xf/BL_Images/Orig_Emoji/wrygrin.gif "wry grin :\")