Phone:
+86-21-69895597
English 
Rm 1955,Building 2,No.215,Lianhe North Rd.,Fengxian Dist.Shanghai 201417 China
Views: 0 Author: Site Editor Publish Time: 2026-11-05 Origin: Site
A quantitative investigation into artificial intelligence applications for reverse-phase high-performance liquid chromatography method development demonstrated that while an AI platform attained an overall prediction accuracy of 72.9 percent for ranolazine analysis, human intervention remained essential to resolve chemical interaction parameters. The research introduced a parameter-level quantitative framework designed to systematically evaluate AI prediction accuracy in RP-HPLC method development, addressing a methodological gap in the existing literature by using ranolazine as a model analyte.
To evaluate predictive capabilities, the researchers tested four separate AI platforms—Perplexity, Copilot, Gemini, and YesChat—using standardized chromatographic inputs. Among these systems, the YesChat HPLC Method Developer platform was selected for in-depth evaluation because it provided structured, actionable outputs along with an explicit focus on green chemistry principles. The research team assessed the platform’s outputs across seven chromatographic parameters scored through a structured three-tier evaluation system.
The evaluation revealed stark differences in prediction performance between physical instrument settings and complex chemical parameters. The AI achieved 100 percent accuracy on four instrumental parameters: column temperature, flow rate, detection wavelength, and mobile phase ratio. In contrast, the mobile phase pH range scored 70 percent because the platform suggested a broad range of 3.0 to 4.5 rather than identifying a specific operational value.
Significant limitations appeared when the AI addressed parameters governing chemical interactions. The model scored 0 percent on buffer necessity and 40 percent on cyclodextrin concentration, demonstrating an inability to correctly determine necessary mobile phase additives. Human-guided optimization subsequently addressed these deficiencies by removing the phosphate buffer entirely and reducing the concentration of sulfobutylether-β-cyclodextrin to 4 millimoles per liter, which improved chromatographic peak shape and retention, resulting in a retention time of 6.17 minutes.
The resulting RP-HPLC method underwent analytical validation in accordance with International Council for Harmonisation ICH Q2(R1) guidelines. Validation confirmed robust performance, exhibiting linear response across a concentration range of 10.0 to 100.0 micrograms per milliliter with a coefficient of determination of 0.9999, precision with a relative standard deviation below 2 percent, and recovery rates ranging from 98.66 to 101.20 percent. The method demonstrated a limit of detection of 1.25 micrograms per milliliter and a limit of quantitation of 3.79 micrograms per milliliter.
In addition to analytical validation, the developed method exhibited favorable environmental performance, achieving an estimated White Analytical Chemistry score of approximately 82 out of 100. This green profile was accomplished by eliminating acetonitrile, decreasing methanol consumption, and utilizing cyclodextrin-assisted optimization. The findings indicate that AI-assisted workflows can substantially accelerate the development of sustainable analytical methods for pharmaceutical quality control, provided human analytical expertise is applied to adjust complex chemical parameters and guarantee method robustness.
