Prediction of the Phase Composition Profile of Three‐Compound Mixtures in Liquid‐Liquid Equilibrium: A Chemoinformatics Approach

Machine‐learning models were developed to predict the composition profile of a three‐compound mixture in liquid‐liquid equilibrium (LLE), given the global composition at certain temperature and pressure. A chemoinformatics approach was explored, based on the MOLMAP technology to encode molecules and...

Full description

Saved in:
Bibliographic Details
Published in:Chemphyschem 2022-12, Vol.23 (24), p.e202200300-n/a
Main Authors: Carrera, Gonçalo V. S. M., Cruz, Mariana L., Klimenko, Kyrylo, Esperança, José M. S. S., Aires‐de‐Sousa, João
Format: Article
Language:English
Subjects:
Atomic properties
big data
Cheminformatics
chemoinformatics
codification
ionic liquid
Ionic liquids
Ionic Liquids - chemistry
Machine learning
Mixtures
Organic chemistry
phase behaviour
Phase composition
Temperature
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Machine‐learning models were developed to predict the composition profile of a three‐compound mixture in liquid‐liquid equilibrium (LLE), given the global composition at certain temperature and pressure. A chemoinformatics approach was explored, based on the MOLMAP technology to encode molecules and mixtures. The chemical systems involved an ionic liquid (IL) and two organic molecules. Two complementary models have been optimized for the IL‐rich and IL‐poor phases. The two global optimized models are highly accurate, and were validated with independent test sets, where combinations of molecule1+molecule2+IL are different from those in the training set. These results highlight the MOLMAP encoding scheme, based on atomic properties to train models that learn relationships between features of complex multi‐component chemical systems and their profile of phase compositions. The MOLMAP encoding method, based on the atomic pattern of activation of a given chemical system on a Kohonen neural network, was used to represent: a‐ a three‐compound mixture (ionic liquid+molecule1+molecule2) and b‐ molecule1 or molecule2 of that mixture at certain temperature and pressure. The Random Forest algorithm predicts its phase behaviour profile.
ISSN:1439-4235
1439-7641
DOI:10.1002/cphc.202200300