When Does Machine Learning Add Value over Theory? Predicting API Solubility in Binary Mixtures with COSMO-RS and DOOIT2 Across Diverse and Homogeneous Systems
Maciej Przybyłek, Tomasz Jeliński, Adrian Drużyński, Piotr Cysewski · 2026
Read When Does Machine Learning Add Value over Theory? Predicting API Solubility in Binary Mixtures with COSMO-RS and DOOIT2 Across Diverse and Homogeneous Systems by Maciej Przybyłek, Tomasz Jeliński, Adrian Drużyński, Piotr Cysewski (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13210068.txt.
Description
Przybyłek and co-authors ask when machine learning beats pure theory in a 2026 chemistry paper. The task is guessing how well drug compounds dissolve in two part solvent mixes, a key step in drug design. A physics based model gives a solid baseline but falters on messy mixes. Their hybrid method adds a learning based fix on top of the theory, pruning features as it goes.
On a uniform set of acid compounds, theory alone already scored well, and the hybrid added almost nothing. The wider set held 85 varied drug compounds and over ten thousand data points, plus fresh measures for three drugs. There the answer differed by data set, as the paper's title promises. The lesson is practical: learning helps most where theory strains, not across the board.
Catalog record
| CREATOR | Maciej Przybyłek, Tomasz Jeliński, Adrian Drużyński, Piotr Cysewski |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 63 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |