Analytical quality by design-based development of a green LC-MS/MS method using a neural network for the simultaneous determination of selected antiviral drugs

An LC-MS/MS method was developed for the simultaneous determination of antiviral drugs (favipiravir, remdesivir, lopinavir, and ritonavir). Critical method parameters were identified and evaluated through preliminary risk assessment within the analytical quality by design framework. The effects of flow rate (0.5–0.7 mL min⁻¹), formic acid content in the mobile phase (0.05–0.10%), cone voltage (15–35 V), and acetonitrile content in mobile phase (65–90%v/v) on critical quality attributes, including retention time, peak area, and resolution were modeled using both multiple linear regression (MLR) and a radial basis function artificial neural network (RBF-ANN). MLR provided insights into factor effects, particularly flow rate and cone voltage, whereas RBF-ANN described complex relationships and enabled the definition of the method operable design region (MODR). The established MODRs were confirmed through multiple verification points. Analytical validation using a simplified protein precipitation protocol demonstrated robust selectivity across six matrix lots, with no detectable carryover. Calibration curves showed strong linearity (R2 > 0.99) over the ranges of 70–700 ng mL−1 for favipiravir and remdesivir, and 80–700 ng mL−1 for ritonavir and lopinavir (R2 > 0.98). Validation results met predefined acceptance criteria, with recoveries of 82–118% and precision values of 1–14% relative standard deviation. The method was applied to pooled clinical plasma samples and exhibited favorable green analytical characteristics based on the RAPI, BAGI, AGREE, GAPI, and the Analytical Eco-Scale assessments. It achieved high sample throughput, reduced solvent consumption, and lower hazardous waste generation.