Accelerated discovery of functional dyes via machine learning and chemical space exploration

Bo Xiao, Naflaa A. Aldawsari, Safaa N. Abdou, Mohamed M. Ibrahim, Asif Mahmood,
Accelerated discovery of functional dyes via machine learning and chemical space exploration,
Solid State Communications,
Volume 406,
2025,
116214,
ISSN 0038-1098,
https://doi.org/10.1016/j.ssc.2025.116214.
(https://www.sciencedirect.com/science/article/pii/S0038109825003898)
Abstract: A novel and efficient framework has been established for designing and screening dyes. Among various machine learning approaches evaluated, gradient boosting regression demonstrated the highest performance and was chosen for further analysis. Using this model, a database of 10,000 dyes was generated, with their UV/visible absorption maxima predicted. Visualization through a t-SNE plot confirmed substantial diversity in both chemical structures and absorption maxima. From this dataset, 30 dyes exhibiting red-shifted absorption were selected for deeper investigation. An evaluation of their synthetic accessibility revealed that half of these candidates had SA scores below 6, suggesting they are relatively easy to synthesize. This framework offers a promising strategy for the rapid and effective screening of dye molecules.
Keywords: Machine learning; Dyes; Light absorption behavior; Feature importance