CatalySeed: A Reaction Database for Ruthenium-Catalyzed Ethenolysis of Seed Oils with Applications in Machine Learning

Ethenolysis of unsaturated
seed oils is an atom-efficient metathesis
reaction that enables α-olefin production and fine chemical
synthesis. By upcycling complex biobased molecules into value-added
products, it supports circular chemical processes. In this study,
we present a curated data set to support machine learning (ML) analysis
of catalytic performance in the ethenolysis of seed oils. Through
a detailed classification of 768 entries and 217 catalysts, along
with the integration of the ROBERT ML framework, with the CatalySeed
database we identify key electronic descriptors that correlate with
experimental outcomes. Binary classification models for TON (threshold
≥ 0.75 × 106) and % selectivity (≥90%)
achieved strong performance, suggesting that higher Ru partial charge
tends to correlate with higher TON, while lower metal d-orbital character
is generally associated with higher selectivity. These findings illustrate
how this database, available through an open-access web server, enables
ML to uncover predictive trends, supporting catalyst design strategies
beyond conventional computational approaches for the transformation
of renewable feedstocks.