Machine learning screening tools for the prediction of extraction yields of pharmaceutical compounds from wastewaters

Aug 4, 2026

Upload by Eliana

Pharmaceutical compounds increasingly contaminate wastewaters, where conventional treatment processes often fail to remove them, leading to environmental discharge. Liquid-liquid extraction can remove these pollutants, while COSMO-RS helps predict molecular interactions and identify promising solvents, but COSMOtherm models cannot capture key process parameters, limiting prediction accuracy. This study applies machine learning to predict extraction yields of eleven pharmaceuticals using eight solvents. Six regression and two classification models are evaluated; ANN regression and RF classification provide the best performance and reveal key yield-controlling descriptors. Machine learning supports effective solvent and condition screening for pharmaceutical removal.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Introduction

Pharmaceutical micropollutants in wastewater

Limitations of conventional treatments

Removal strategies (biological, oxidation, separation)

Liquid-liquid extraction and solvent concerns

Alternative solvents: eutectic solvents and terpenoids

Need for computational solvent screening

COSM-RS/COSMO-RS background and limitations

Study motivation for ML-based prediction

Journal of Water Process Engineering 62 (2024) 105379 Contents lists available at ScienceDirect Journal of Water Process Engineering journal [homepage: www.elsevier.com/locate/jwpe](homepage: www.elsevier.com/locate/jwpe) | Machine learning screening tools for the prediction of extraction yields of pharmaceutical compounds from wastewaters
Ana Casasa, Diego Rodríguez-Llorente b, *, Guillermo Rodríguez-Llorente c, Juan Garcíab, Marcos Larriba b
a School of Water, Energy and Environment, Cranfield University, Cranfield MK43 0AL, UK
b Catalysis and Separation Processes Research Group (CyPS), Department of Chemical Engineering and Materials, Complutense University of Madrid, Avda. Complutense s/n, 28040 Madrid, Spain
c Department of Artificial Intelligence, HI Iberia, C. de Juan Hurtado de Mendoza, 14, 28036 Madrid, Spain | | | | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | Editor: Ludovic F. Dum´ee | | Pharmaceutical compounds have become an increasingly important source of pollutants in wastewaters being conventional treatments ineffective in removing them, so they are commonly discharged into the environment. Pharmaceuticals can be successfully removed using liquid-liquid extraction, and COSMO-RS can be used to predict interactions and identify the most promising solvents. However, COSMOtherm models cannot account for key process parameters, which reduces the accuracy of these computational models. Therefore, there is a need for alternative computational approaches to accurately predict the extraction yields of pharmaceuticals which can incorporate both processing and interaction variables. This work used machine learning to predict the extraction yield of eleven pharmaceuticals using eight solvents. Six regression models and two classification models were explored. The best performance was obtained with ANN regressor (test MAE: 4.510, test R2: 0.884) and RFclassifier (test accuracy: 0.938, test recall: 0.974). The RF regression analysis and classification also showed key extraction yield features: solvent-to-feed ratio, n–octanol–water partition coefficient, hydrogen bond and Van der Waals contributions to excess enthalpy, and pH distance to nearest pKa. Machine learning showed as an excellent tool for screening and selecting the most promising solvents and process conditions to remove pharmaceuticals from wastewater. | | Keywords:
Machine learning COSMO-RS
Liquid-liquid extraction Pharmaceuticals Wastewater | | | 1. Introduction Pharmaceuticals are a significant source of micropollutants in wastewater, with hospital wastewater being a major contributor [1,2]. Long-term exposure to pharmaceuticals in trace amounts can harm aquatic ecosystems and potentially impact human health. These substances enter sewage treatment plants but are not effectively removed by conventional treatment processes, leading to their discharge into the natural environment [3]. The presence of pharmaceuticals in water environments raises concerns as antimicrobial resistance and endocrine disruption effects [4,5]. Various techniques are being explored to eliminate pharmaceuticals from wastewater, including biological processes [6,7], oxidation methods [8–12], and separation processes [13–15]. Liquid-liquid extraction is a well-established method, but using hazardous conventional solvents is a concern. To address this, alternative and more sustainable solvents like natural eutectic solvents [16–21], and terpenoids [22,23] are being explored. Eutectic solvents and terpenoids have been shown in literature as promising solvents in the extraction of pharmaceutical compounds from wastewater [24]. However, there is an unlimited number of combinations of cations, anions and hydrogen bond donors and acceptors that can be used to constitute eutectic solvents, and the literature on the use of terpenoids in this context is scarce. For those reasons and because laboratory testing is expensive and time consum

FAQs

Why are pharmaceutical compounds a concern in wastewater?

Pharmaceuticals act as micropollutants and persist in trace concentrations, potentially harming aquatic ecosystems and affecting human health. They enter sewage treatment plants but are not effectively removed by conventional processes, leading to environmental release.

How does COSMO-RS relate to solvent selection for extraction?

COSMO-RS predicts interactions and thermodynamic properties to support screening of solvents for liquid-liquid extraction. It can guide selection, but COSMOtherm-based approaches may not account for important process parameters, reducing overall prediction accuracy.

What machine learning approach was used to predict extraction yields?

The work uses machine learning to predict extraction yields of eleven pharmaceuticals from eight solvents, testing six regression models and two classification models. Best results were obtained with an ANN regressor and an RF classifier, and the analysis identifies key features influencing extraction yield.

Aug 4, 2026

Upload by Eliana