Shujun He

This paper demonstrates that deep learning can accurately design RNA pseudoknots, achieving results comparable to human experts in a competition.

Struct2SeQ uses deep Q-learning to design RNA sequences that fold into target structures and match SHAPE reactivity constraints, outperforming humans and existing methods.

The paper describes a Kaggle competition for RNA 3D structure prediction, showing that template-based methods can outperform deep learning approaches.

Across 42 CASP16 nucleic acid targets, accuracy stayed tied to template availability; human experts (Vfold first) led servers and AlphaFold 3, though servers now improve on templates on average.

The paper introduces RibonanzaNet, a deep learning model trained on crowdsourced RNA chemical mapping data to predict RNA structure and related properties.

The paper introduces the Nucleic Transformer, a deep learning model for DNA/RNA tasks that is effective, interpretable, and performs well with limited data in promoter classification, viral genome identification, and mRNA vaccine degradation analysis.