Shiyi Du

CodonRL improves codon sequence optimization using reinforcement learning with structural feedback and expert demonstration, outperforming existing methods on multiple metrics.

ARCADE enables flexible, controllable codon design using activation engineering from genomic language models to steer biological metrics like Codon Adaptation Index and GC content.

The paper evaluates index mis-assignment rates on DNB-based sequencing platforms and finds they are exceptionally low, suggesting reliable multiplex sequencing.