Prediction of cellulose dissolution in ionic liquids using molecular descriptors based QSAR model

The dissolution of lignocellulose by ionic liquids attracted much attention during the last decade. However, the experimental screening and selection of a large number of potential ionic liquids for biomass processing are challenging task. In this study, the prediction of cellulose dissolution in ionic liquids was evaluated by quantitative structure-activity relationship (QSAR) model using the molecular descriptors of ionic liquid's constitutional ions derived from CODESSA program. Models based on cellulose molar solubility exhibited better correlation (R2 of 0.93 vs 0.88) and predictability (R2 of 0.89 vs 0.83) than those based on mass percentage solubility. In addition, models developed by multivariate adaptive regression spline (MARS) method employed less variables (13 vs 57-59) and showed better predictability (R2 of 0.83-0.89 vs 0.45-0.51) compared to those developed by multiple linear regression (MLR) technique. The results indicated that the molecular descriptor of ILs could be effectively used to develop QSAR models for facilitating the in silico and a priori screening/selection of ILs customized for specific applications.