The aim of this paper is to recommend a new strategy for the analytical validation based on the uncertainty profile as a graphical decision-making tool, and to exemplify a novel method to estimate the measurement uncertainty. Indeed, the innovative formula that we offer to assess the uncertainty is based on the calculation of the β-content tolerance interval. Three chemometric methodologies are exposed to build the (β, γ) tolerance interval, namely: the Satterthwaite approximation, the GPQ method (generalized pivotal confidence) and the MLS procedure (modified large simple). Furthermore, we illustrate the applicability and flexibility of the uncertainty profile to assess the fitness of the purpose of chromatographic and electrophoretic analytical methods, which use different instrumental techniques such as liquid chromatography (LC-UV, LC-MS), gas chromatography (GC-FID, GC-MS) and capillary electrophoresis (CE, CE-MS). In addition, we demonstrate here that (β, γ) tolerance intervals will provide perfect estimates of the routine uncertainty. In particular, we show that there is no difference statistically between the uncertainties estimated by our methodology as of the validation stage with those obtained from the routine phase.

Multidrug resistance (MDR) proteins related to the ATP-binding cassette family are found in a very wide range of human tumors and result in therapeutic failure. The overexpression of efflux pumps such as ABCB1 is one of the mechanisms of MDR. This paper aims to develop a reliable quantitative structure-activity relationship (QSAR) model that best describes the correlation between the activity and the molecular structures in order to predict the inhibitory biological activity towards ABCB1. In this regard, a series of quinoline derivatives of 18 compounds were analyzed using different linear and non-linear machine learning (ML) regression methods including k-nearest neighbors (KNN), decision tree (DT), back propagation neural networks (BPNN) and gradient boosting-based (GB) methods. Their aim is to explain the origin of the activity of these investigated compounds and therefore, design new quinoline derivatives with higher effect on ABCB1. A total of 16 ML predictive models were developed on different number of 2D and 3D descriptors and were evaluated using the coefficient of determination (R²) and the root mean squared error (RMSE) statistical metrics. Among all developed models, A GB-based model in particular catboost achieved the highest predictive quality, with one descriptor, expressed by R² and RMSE of 95% and 0.283 respectively. Molecular docking studies against the target crystal structure of the outward-facing p-glycoprotein (6C0V) revealed significant binding affinities via both hydrophobic and H-bond interactions with the relevant compounds. The 17 has shown the highest binding energy of -9.22 kcal/mol. Therefore, it can suggest that 17 may prove to be a valuable potential lead structure for the design and synthesis of more potent P-glycoprotein inhibitors for combination used with anti-cancer drugs for cancer multidrug resistance management.