The Price is Right: Predicting Reagent Prices

18 March 2021, Version 1
This content is a preprint and has not undergone peer review at the time of posting.

Abstract

We present a model for estimating the price of a reagent from its chemical structure. It is intended to be useful when doing reagent selection for library design. The model is a Random Forest regressor which is trained on the MolPort catalog of 302K reagents and the log of their price. For descriptors we use topological fingerprints from RDKit: chiral Morgan fingerprints, its medicinal chemistry descriptors, and counts of undetermined chiral centers. The model has an out-of-bag performance of 34% variance explained in log Price. When predicting on known reagents, the model explains 91% of the variance in log Price. We analyzed the model to understand the errors that the model makes. We show that the compounds with the highest errors have only a subtly different structure from similar molecules, but very different in price.

Keywords

Reagent Pricing
Machine learning
Random forest
Cliff pairs
Matched molecular pairs

Comments

Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.