Ali Kaiyal presented his research in a faculty seminar!

In his seminar, Ali, an MSc student, shared his work on deep learning-based classification of single proteins using nanochannel fluorescence sensing data. He presented a novel machine learning framework that analyzes the rich optical and physical trajectories generated as individual proteins pass through a nanochannel, enabling direct protein identification from raw sensing data.

Ali’s approach is built around a custom hybrid neural network architecture, MaskedInceptionTST, designed to handle variable-length sequences, class imbalance, and cross-session generalization. The model also addresses the challenge of identifying previously unseen protein classes by recognizing when an input does not belong to any class encountered during training.

This work lays the foundation for intelligent, data-driven single-molecule protein identification, with strong potential for fast, accurate, and noninvasive sensing platforms for future biomedical applications.

Great work, Ali!