Transfer Learning\nApproach to Multitarget QSRR Modeling\nin RPLC

QSRR is a valuable technique for the retention time predictions\nof small molecules. This aims to bridge the gap between molecular\nstructure and chromatographic behavior, offering invaluable insights\nfor analytical chemistry. Given the challenge of simultaneous target\nprediction with variable experimental conditions and the scarcity\nof comprehensive data sets for such predictive modelings in chromatography,\nthis study introduces a transfer learning-based multitarget QSRR approach\nto enhance retention time prediction. Through a comparative study\nof four models, both with and without the transfer learning approach,\nthe performance of both single and multitarget QSRR was evaluated\nbased on Mean Squared Error (MSE) and <i>R</i><sup>2</sup> metrics. Individual models were also tested for their performance\nagainst benchmark studies in this field. The findings suggest that\ntransfer learning based multitarget models exhibit potential for enhanced\naccuracy in predicting retention times of small molecules, presenting\na promising avenue for QSRR modeling. These models will be highly\nbeneficial for optimizing experimental conditions in method development\nby better retention time predictions in Reversed-Phase Liquid Chromatography\n(RPLC). The reliable and effective predictive capabilities of these\nmodels make them valuable tools for pharmaceutical research and development\nendeavors.