MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641083261 A) filed by Manjesh Mathew; and Dr. Renjith Thomas on July 07, 2026, for Machine Learning Model Based Prediction System For Gas-Phase Proton Affinity.
Inventors include Manjesh Mathew; and Dr. Renjith Thomas.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: A ML model predicting the gas-phase PA of a molecule directly from its molecular structure to identify molecules exhibiting superbasic or proton-sponge character. A molecular structure, received as a SMILES, converted into a 3D conformer, which a composite feature vector is generated 2D physicochemical descriptors, Morgan- type and MACCS-type fingerprints, 3D shape descriptors, and a chemist-curated substructure flags encoding protonsponge pharmacophores. The composite feature vector is supplied to an ensemble of trained I regression models, the outputs of which are combined by voting or stacking aggregator, to generate a predicted proton- affinity value. The predicted value is compared against a threshold to classify the molecule, and a feature-attribution module generates per-feature contribution scores explaining the structural basis of the prediction. The disclosed method reduces the cost and time with quantum-chemical-proton-affinity determination by orders of magnitude. retaining chemically interpretable outputs, and is suited to virtual screening of candidate superbases, proton sponges, and organocatalysts.
Disclaimer: Curated by HT Syndication.
