Predicting Kováts Retention Indices Using Graph Neural Networks

The Kováts retention index is a dimensionless quantity that characterizes the rate at which compound processed through gas chromatography column. This independent of many experimental variables and, as such, considered near-universal descriptor time on indices large number molecules have been determined experimentally. “NIST 20: GC Method/Retention Index Library” database has collected more importantly, curated subset these compounds resulting in highly valued reference database. data library form an ideal set for training machine learning models prediction unknown compounds. In this article, we describe graph neural network model to predict NIST and compare approach with previous work [1]. We mean unsigned error 28 units compared 44, putative best result using convolutional also incorporates estimation scheme based group contribution achieves 114 data. Our method uses same input source approach, making its application straightforward convenient apply existing libraries. results convincingly demonstrate predictive powers systematic, data-driven approaches leveraging deep methodologies applied chemical 20 outperform models.