# Toward AI-Assisted Greener Chiral HPLC: Predicting Efficient Enantioseparation–Mobile Phase (EES–MP) Profiles for MP SelectionA Lux Cellulose-1 Case Study

## TL;DR

This paper introduces an AI model to predict optimal mobile phases for chiral HPLC, reducing trial-and-error and making the process greener and faster.

## Contribution

The novel contribution is the first proof-of-concept for in silico prediction of full EES–MP profiles in chiral HPLC using a consensus ANN model.

## Key findings

- The ANN-consensus model accurately reproduces EES–MP profiles with R² > 0.9 and low error dispersion.
- External tests on fluoxetine and lormetazepam confirmed the model's ability to anticipate separability and MP selection.
- The framework offers a greener, data-efficient approach to chiral HPLC method development.

## Abstract

Chiral HPLC method development still relies heavily on
trial-and-error
screening. We introduce the Efficient Enantioseparation (EES) parametera
single metric integrating resolution (Rs) and retention (k)to
move from point predictions to full mobile-phase (MP) profile modeling.
Using EES as the response, we trained multiple artificial neural networks
(ANNs) on 62 variables (molecular descriptors) and 76 objects (structurally
diverse neutral and basic compounds chromatographed on a Lux Cellulose-1
column under aqueous–acetonitrile conditions at nine MP compositions).
ANNs were optimized with a chaotic competitive-learning optimizer
(CCLNNA), then ranked/selected and combined into a consensus model
to enhance robustness and limit overfitting. The ANN-consensus model
accurately reproduces full EES–MP profiles (R
2 > 0.9) with lower error dispersion, enabling prospective
feasibility checks and single-shot selection of high-EES mobile-phase
compositions. External tests on fluoxetine and lormetazepam confirmed
prospective utility by anticipating separability at one or more MPs
(nominating the MP with maximal EES) or nonseparability across the
explored MP range. To our knowledge, this work provides the first
proof-of-concept for in silico prediction of full EES–MP profiles
in chiral HPLC, enabling intelligent MP selection. Rather than a definitive
model, this work evaluates the potential of the strategy: consensus
stabilizes learning with limited data and offers greener, actionable
guidance that can reduce experiments, reagent consumption, and development
time. The framework is extensible to broader chemotypes and stationary
and mobile phases; larger data sets could further generalize EES-profile
prediction and support intelligent MP selection in sustainable chiral
HPLC.