Fluor-Predictor: An Interpretable Tool for Multiproperty Prediction and Retrieval of Fluorescent Dyes DOI
Wenxiang Song,

Le Xiong,

Xinmin Li

et al.

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: March 12, 2025

With the rapid advancements in field of fluorescent dyes, accurate prediction optical properties and efficient retrieval dye-related data are essential for effective dye design. However, there is a lack tools comprehensive integration convenient retrieval. Moreover, existing models mainly focus on single property dyes fail to account diverse fluorophores solutions systematic manner. To address this, we proposed Fluor-predictor, multitask model fluorophores. This study integrates multiple databases develops an interpretable graph neural network-based regression predict four key dyes. We thoroughly examined impact factors such as quality number solvents performance. By leveraging atomic weight contributions, not only predicts these but also provides insights guide structural modifications. In addition, compiled built database containing 36,756 records fluorescence properties. limitations Xanthene Cyanine then 1148 1496 from literature, comparing direct training with transfer learning approaches. The achieved mean absolute errors (MAE) 11.70 nm, 15.37 0.096, 0.091 predicting absorption wavelength (λabs), emission (λem), quantum yield (Φ) molar extinction coefficient (Log(ε)), respectively. integrated this work into tool, which supports methods multiproperty prediction. Fluor-predictor will facilitate retrieval, prescreening, modification

Language: Английский

Fluor-Predictor: An Interpretable Tool for Multiproperty Prediction and Retrieval of Fluorescent Dyes DOI
Wenxiang Song,

Le Xiong,

Xinmin Li

et al.

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: March 12, 2025

With the rapid advancements in field of fluorescent dyes, accurate prediction optical properties and efficient retrieval dye-related data are essential for effective dye design. However, there is a lack tools comprehensive integration convenient retrieval. Moreover, existing models mainly focus on single property dyes fail to account diverse fluorophores solutions systematic manner. To address this, we proposed Fluor-predictor, multitask model fluorophores. This study integrates multiple databases develops an interpretable graph neural network-based regression predict four key dyes. We thoroughly examined impact factors such as quality number solvents performance. By leveraging atomic weight contributions, not only predicts these but also provides insights guide structural modifications. In addition, compiled built database containing 36,756 records fluorescence properties. limitations Xanthene Cyanine then 1148 1496 from literature, comparing direct training with transfer learning approaches. The achieved mean absolute errors (MAE) 11.70 nm, 15.37 0.096, 0.091 predicting absorption wavelength (λabs), emission (λem), quantum yield (Φ) molar extinction coefficient (Log(ε)), respectively. integrated this work into tool, which supports methods multiproperty prediction. Fluor-predictor will facilitate retrieval, prescreening, modification

Language: Английский

Citations

0