QSPR-driven prediction of analyte permeability for advancing CP-MIMS applications

Condensed Phase-Membrane Introduction Mass Spectrometry (CP-MIMS) is a sustainable and highly versatile approach within the framework of Direct Mass Spectrometry techniques, which enables real-time determination of target analytes by integrating sampling, purification, and mass-spectrometric analysis into a single step. Liquid or slurry samples are directly coupled to a mass spectrometer through a semipermeable membrane, with transport of analytes governed by their physicochemical and structural properties. At present, experimental testing is often required to determine whether a compound is suitable for CP-MIMS analysis, limiting the scalability of this approach. To overcome this limitation, computational strategies, such as Quantitative Structure-Property Relationship (QSPR), offer the possibility to rationalize and predict analyte behavior at the membrane interface, enabling a rapid recognition of permeant compounds. We developed a QSPR classifier to predict whether a compound can permeate a polydimethylsiloxane membrane in the CP-MIMS system by using machine learning methods to relate molecular structural features to physicochemical properties. To improve robustness, the literature derived training set was expanded through experimental measurements and similarity searches in large molecular databases. The resulting model achieved high sensitivity and specificity in large-scale external validation. Experimental testing of QSPR-predicted permeant candidates confirmed the reliability of the model predictions. Moreover, a further experimental evaluation was performed on environmentally relevant contaminants under CP-MIMS conditions. The proposed QSPR framework provides a reliable and automated in-silico tool for pre-screening of permeant analytes, supporting more efficient method development and improved sustainability for advancing CP-MIMS applications in several fields.