Developing a machine learning model for accurate nucleoside hydrogels prediction based on descriptors.

Area of Science:

• Biomaterials Science
• Supramolecular Chemistry
• Computational Chemistry

Background:

• Nucleoside-derived supramolecular hydrogels show promise in biomedicine due to biocompatibility.
• Predicting hydrogel formation from nucleoside derivatives remains a significant challenge.
• Lack of predictive models hinders the discovery of new hydrogel materials.

Purpose of the Study:

• To develop a machine learning model for predicting the hydrogel-forming ability of nucleoside derivatives.
• To identify novel nucleoside derivatives with hydrogelation capabilities.
• To explore potential applications of newly discovered hydrogels.

Main Methods:

• Machine learning model development using a dataset of 71 nucleoside derivatives.
• Model optimization and validation to achieve 71% accuracy.
• Experimental verification of hydrogel formation for selected molecules.

Main Results:

• An accurate machine learning model (71% accuracy) was established for predicting hydrogel formation.
• 24 nucleoside derivatives were computationally screened, and their hydrogel-forming ability was experimentally confirmed.
• Two novel cation-independent nucleoside hydrogels were discovered and characterized.
• The cation-independent hydrogels demonstrated potential for detecting Ag+ and cysteine.

Conclusions:

• Machine learning provides a valuable tool for predicting nucleoside derivative hydrogelation.
• The discovered cation-independent hydrogels offer new possibilities for sensing applications.
• This approach accelerates the discovery and development of functional supramolecular hydrogels.