What question did this study set out to answer?

March 17, 2026Open Access

Experimental Multivariate Modeling of Structure-Fluorescence Relationships in Organic Compounds under Chromatographic Conditions: A Proof-of-Concept Study

Key Points

The aim is to investigate structure-fluorescence relationships in organic compounds using advanced computational methods.
Used a dataset of 117 observations analyzed by Reverse Phase High Pressure Liquid Chromatography with Fluorescence Detection.
Employed Partial Least Squares (PLS) and Artificial Neural Network (ANN) for modeling fluorescence behavior.
Included 36 physicochemical properties and 38 structural descriptors as predictor variables.
Identified key structural features influencing fluorescence, such as fused ring systems and functional groups.
Physicochemical properties like lipophilicity and molecular flatness showed significant roles.
ANN predictions for unknown molecules achieved high reliability with correlation coefficients greater than 0.85.

Abstract

• Computational fluorescence study using 117 observations and 74 descriptors • PLS and ANN multivariate modeling for the elucidation of fluorescence • In silico models hierarchically classified key fluorescence factors. Although fluorescence provides valuable analytical information, its theoretical interpretation remains challenging. In silico methodologies offer a systematic framework to explore multifactorial mechanisms and enhance fluorometric predictions. The present study investigates theoretical aspects of fluorimetry by interpreting the behavior of 117 substances analyzed by RP‑HPLC‑FLD (Reverse Phase High Pressure Liquid Chromatography with Fluorescence Detection), in methanol both as diluent and mobile phase, using Partial Least Squares (PLS) and Artificial Neural Network (ANN) models. The dataset used for model development consisted of the logarithmically transformed fluorescence responses of 117 observations (Y variable), while 36 physicochemical properties and 38 structural descriptors were included as predictor variables (X). PLS analysis objectively quantified the relative importance of the factors governing fluorescence, revealing that dominant contributors were primarily structural features such as fused or aromatic ring systems (+), conjugated single-double bonds (+), carbonyl groups (-), carboxylate groups (-), hydroxyl groups (+), amino groups (+), and sulfur-containing moieties (-). Several physicochemical properties also played a significant role, including lipophilicity (+), polar surface area (-), molecular flatness (+), hydrogen‑bond donor/acceptor capacity (-), and molecular flexibility (-). Accordingly, the application of ANN enabled the prediction of fluorescence intensity for unknown molecules with high reliability (r > 0.85). It was also noteworthy that the signal intensity of an analyte was influenced by the mobile phase flow rate as a function of the slope of its calibration curve. Overall, this multivariate approach provides, for the first time, a clear and objective prioritization of the dominant descriptors shaping fluorometric behavior.

Read Full Paperexternally

Mark Helpful

Bookmark

Relay

View Full Paper