Screening of Natural Anti-Aging Compounds Based on WBDSS Algorithm and Transfer Learning

This article addresses the challenge of traditional sampling algorithms struggling to generate representative samples of minority classes by proposing a dual-safe mixed sampling algorithm based on Wasserstein distance (WBDSS). The aim is to tackle the issue of sample imbalance in the screening of natural anti-aging compounds. WBDSS combines Wasserstein distance with a dual safety mechanism to generate high-quality minority class samples by improving the SMOTE algorithm, and fine-tunes the pre-trained TabPFN model to enhance the predictive accuracy in the screening of anti-aging compounds. Experimental results indicate that WBDSS outperforms traditional sampling algorithms across multiple evaluation metrics, and the fine-tuned TabPFN model, enhanced by WBDSS data augmentation, performs well in the task of screening natural anti-aging compounds, effectively identifying natural anti-aging agents.