Multiple Molecule + "concentration" Regression Task

[kamilm]

Hello,

First, thank you all for the work and open sourcing such a robust and powerful model architecture/package.

I have a regression task where I my inputs are molecules and concentrations, I am curious if this is within the scope of ChemProp? and if so, where in the docs can I read about implementing this? I'm not above misusing the package and
hacking something together, I'm just not sure where to look in the package so that I can couple "molecules" and a "property" 😄 .

My data and problem statement are as such:

molecule-A @ conc-A1 + molecule-B @ conc-B1 + molecule-C @ conc-C1 gives result-1
molecule-A @ conc-A2 + molecule-B @ conc-B2 + molecule-C @ conc-C2 gives result-2
molecule-A @ conc-A3 + molecule-B @ conc-B3 + molecule-C @ conc-C3 gives result-3
...
molecule-D @ conc-D1 + molecule-E @ conc-E1 + molecule-F @ conc-F1 gives result-M
...
molecule-X @ conc-Xn + molecule-Y @ conc-Yn + molecule-Z @ conc-Zn gives result-N

I would like to be able to pass in:

new-molecule-A @ new-conc-A + new-molecule-B @ new-conc-B + new-molecule-C @ new-conc-C

and have the output be some continuous value.

Thank You, I hope this was clear. I can provide further clarification if needed.

-Kamil

[kevingreenman]

Hi @kamilm, apologies for the slow reply. This functionality is not currently available in any version of chemprop, but we have been working on projects related to mixed solvents and areas like that where this would be useful, and this may become a part of chemprop in the future.