Artificial intelligence in chromatography: Greenness and performance evaluation of AI-predicted and in-lab optimized HPLC methods for simultaneous separation of amlodipine, hydrochlorothiazide, and candesartan

Abstract

Artificial Intelligence (AI) is revolutionizing analytical chemistry by promising rapid method development, yet its real-world efficacy remains untested in complex pharmaceutical separations. This study leverages AI to design an HPLC method for Amlodipine (AMD), Hydrochlorothiazide (HYD), and Candesartan (CND), comparing it with an experimentally optimized approach to uncover practical benefits and limitations. Our findings reveal AI\"s potential to accelerate innovation while highlighting the critical role of human expertise. The In-Lab optimized HPLC method utilized an Xselect CSH Phenyl Hexyl® (2.5 µm, 4.6 × 150 mm) column with a mobile phase of acetonitrile:water (0.1 % trifluoroacetic acid) (70:30, v/v), a flow rate of 1.3 mL/min, and UV detection at 250 nm. It achieved rapid elution with retention times of AMD = 0.95 min, HYD = 1.36 min, and CND = 2.82 min. The AI-generated method used a C18 column (5 µm, 150 mm × 4.6 mm), gradient elution with phosphate buffer (pH 3.0) and acetonitrile, a flow rate of 1.0 mL/min, and detection at 240 nm, resulting in longer retention times: AMD = 7.12 min, HYD = 3.98 min, and CND = 12.12 min. Linearity ranges were AMD (25.0–250.0 µg/mL), HYD (31.2–287.0µg/mL), and CND (40.0–340.0µg/mL) for the In-Lab method, and AMD (30.0–250.0µg/mL), HYD (35.0–285.0µg/mL), and CND (50.0–340.0 µg/mL) for the AI-HPLC method. Both approaches were validated per ICH guidelines, confirming specificity, accuracy, and reliability. The obtained results were statistically compared with the reported ones using the F-test and Student\"s t-test. In terms of sustainability, the In-Lab method outperformed the AI-based method according to MoGAPI, AGREE, and BAGI assessments, due to reduced solvent use, waste generation, and analysis time. This study underscores the necessity of human intervention to refine AI-generated methods, aligning them with both analytical efficiency and green chemistry goals. Improving AI tools to predict optimal HPLC conditions is essential for advancing sustainable and effective analytical practices.