mattodd
commented
9 months ago Hi @ucnvwqt @qxsml - hopefully we can get together pretty soon to discuss what's remaining now that the original MPharm project that kick-started this project is complete. Let's find a time and go through what would be next.
One thing that came up recently in in-house discussions was whether control compounds are used in the assays that measure the solubility of the compounds we've been looking at. i.e. are there lit solubility standards that are included each time these measurements are made where the compounds have widely-accepted, easily-repeated experimental solubility? At the moment we are rather damning of the software and faithful in the assay results, but we ought to not necessarily assume that the assay data are perfect. (and that's even before addressing whether the software is predicting thermodynamic, rather than kinetic, solubility...)
Hi @ucnvwqt - OK so two things I think are crucial at this point (where we're amazed that the correlation between calculated and measured solubility is so baaad).
1) Stoplight paper. There's a dataset in there that was used to derive the algorithm. We need that plotted as you have done. Is the corrleation there still bad?
2) Literature, more generally. We need to know who else has looked at this, and what they found. Ideally you should assemble such things on the wiki here, but you can use this issue to dump things if easier. Some wider reading might be found in the blog post I've linked to in the wiki section (why don't we reach out to the author of that piece, once we've done our homework?)
Christmas comes first, but if you're bored and in need of some science, I would do the above.