1
$\begingroup$

In the paper Junction Tree Variational Autoencoder for Molecular Graph Generation, the authors propose a ML model that is based on the message passing neural network, variational auto-encoder, the tree decomposition or loopy belief propagation algorithm, to design a ML model that is able to generate molecules with high potency.

What is a molecule with high potency? How can we measure the potency of a molecule? How do we compare two molecules in terms of potency?

|improve this question
$\endgroup$
  • $\begingroup$ I've opened all three PDFs and I couldn't find a single mention of "potency". Where does it show up and in what context? $\endgroup$ – andselisk♦ May 11 '19 at 18:27
  • 2
    $\begingroup$ You probably mean this: en.wikipedia.org/wiki/Potency_(pharmacology) More generally you can characterize potency by the concentration at which an activity is first observed. $\endgroup$ – Buck Thorn May 11 '19 at 18:32
  • 1
    $\begingroup$ @andselisk One of the authors mentions it in the following presentation: Junction Tree Variational Autoencoder for Molecular Graph Generation. I haven't fully read the main JT-VAE paper, so they are likely using a different term to describe it. $\endgroup$ – nbro May 11 '19 at 18:40
  • $\begingroup$ @NightWriter I more or less understood that potency is a relative property of a molecule that measures the amount of that drug required to obtain a certain effect. However, I also have other questions. I would like to understand how potency could be measured in the context of the paper JT-VAE. $\endgroup$ – nbro May 11 '19 at 18:44
  • 1
    $\begingroup$ In "VAE" they used logP (octanol-water partition coefficient) as a dummy property to be optimized, presumably since it's a readily predicted property. It was an entirely computational study so they didn't test any compounds for actual pharma potency. $\endgroup$ – Buck Thorn May 11 '19 at 18:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.