Idea and design
Protein expression and gene expression levels in tobacco leaves infected by viral vectors can be highly variable. We aimed to create an AI model using large datasets, but halfway through the project, we found that the available data on tobacco in the database was limited, and data on protein expression regions and locations was even more scarce. Despite these challenges, we maintained a learning mindset and completed the entire AI training process. We used only 300 tobacco-related entries from UniProtKB for training, employing a PyTorch-based multi-label classifier. If a comprehensive database on viral infections in tobacco can be established in the future, it will lead to a better and more complete understanding of expression levels and vector design.