Author(s)

Yanzhe Wang

Corresponding Author

Yanzhe Wang

Abstract

Anxiety disorders affect over 300 million people worldwide, highlighting the urgent need for efficient drug discovery. Current treatments target neurotransmitter transporters, but traditional. development is slow and costly. We propose a fully connected neural network framework that integrates molecular descriptors from small molecules (via Mordred) with structural features of transporters predicted by AlphaFold. This multimodal representation enables accurate prediction of drug-transporter binding affinity. Experiments show that our model outperforms. classical machine learning baselines and that combining ligand and protein features yields the best results. The framework provides a scalable tool for screening candidate anxiolytic drugs potentially accelerating discovery and reducing trial-and-error costs.

Keywords

Unsupervised learning; Multimodal representation; Mind-body state modeling