Adaptive Conformer Sampling for Property Prediction Using the Conductor-like Screening Model for Real Solvents

Abstract

The valorization of lignocellulose-derived bioproducts requires effective separation from excessive water. Liquid−liquid extraction is a promising low-energy separation technology, but effective extraction requires solvent selection based on the thermodynamic properties of the bioproduct and solvent components. We propose a computational framework for predicting such properties by developing an adaptive conformer selection approach for use with COSMO-RS (conductor-like screening model for real solvents) calculations. In this framework, molecular dynamics simulations are used to generate many molecular structures (conformers) at representative temperatures in varying solvent environments. Conformers are then clustered based on structural metrics in a low-dimensional space and selected using a mixed-integer quadratic programming problem to iteratively insert a sampled conformer. At each iteration, we determine bioproduct properties using COSMO-RS. We demonstrate the capability of the proposed framework on representative bioproducts to show convergence of the adaptive sampling toward experimentally measured properties with fewer calculations than required by random conformer sampling, enabling the improved screening of solvent systems for liquid-phase separation.

Heuristic Computational Model for Predicting Lignin Solubility in Tailored Organic Solvents

“…Conformer selection has been shown to substantially impact COSMO-RS predictions, particularly for flexible molecules like the oligomers studied here. 32,45 Therefore, we selected conformers based on structures identified in good or poor solvents as feasible extremes expected to be observed in realistic solvent systems; a similar strategy utilized in prior work by our group has led to predictions of polymer solubility in good agreement with experiments. 32 To quantify the conformer's structure, the solvent-accessible surface area (SASA) and radius of gyration (R g ) of each pentamer were recorded in each 10 ps time interval for the last 4 ns of each production run by using the MDTraj library.…”

An Explainable Classification Framework for Determining and Understanding the Suitability of Solvent Extraction for Bioproduct Recovery

“…The details of property prediction , are given in Section S2 in the Supporting Information. For separation modeling, we consider and discuss three representative extraction structures based on the ranking of relative volatility (Section ).…”

Large-scale computational polymer solubility predictions and applications to dissolution-based plastic recycling

“…Sampling conformers that cover a range values of these two parameters can provide reliable input for COSMO-RS solubility estimations. 33,39 We thus selected conformers by superimposing a square grid over the two-dimensional R g -SASA scatter plot and choosing conformers closest to grid intersections following our previous work. 33 The number of sampled conformers for each polymer is listed in Table 1.…”