Jean-Claude Bradley

Thanks Egon - without the CDK I don't think it would have been possible to keep the modeling component Open. Are there any descriptors in the CDK that can help us with the symmetry component (i.e. t-butanol is more symmetric than 1-butanol, etc.)? - Jean-Claude Bradley

Egon - regarding your comment on the blog, at this point there really isn't a legal, financial or technical reason to not flag melting points in publications as they go through the peer-review process. If publishers don't adopt these tools, any reviewer can now do so for chemistry papers reporting melting points in characterization data or papers that do modeling of melting points. - Jean-Claude Bradley

Absolutely true! We should see to it that tools that allow to do such are mentioned in the journals' notes to the reviewers! - Egon Willighagen

Regarding the descriptor, I think the Wiener descriptor's are designed to do just that: capture the amount of branching. But, the problem is that the modeling is eager on finding general patterns, and overall this descriptor might not be found as significant... this is one of the challenges of 'molecular chemometrics' :) - Egon Willighagen

Something that identifies the number of equivalence classes would be an indicator of symmetry. I think the WHIM or one of its variants does that or something similar - Rajarshi Guha

Thanks Rajarshi - we'll take a look at WHIM. - Jean-Claude Bradley

Egon - I think (like solubility) we'll end up with several different models depending on the training set. Even now our Open melting point collection does not include salts, inorganics and organometallics. Just doing a quick analysis, we can see that our model underestimates the mp of adamantane and 1-hydroxyadamantane as expected because symmetry consideration is missing. However, it... more... - Jean-Claude Bradley