Kaiyu Nie
I frame RNA design as optimization and machine-learning problems, and build reproducible scientific software to solve them. Finishing my MASc at Concordia on RNA inverse folding.
Featured project
MASc thesisA Modular Multi-Tool Evolutionary Framework for RNA Inverse Folding
CEC 2026 · OralA predictor-agnostic island-model evolutionary algorithm designed from scratch — population structure, fitness formulation, and mutation operators. A multi-oracle consensus fitness combines ViennaRNA and MXFold2 to make design robust under predictor uncertainty, reaching a 68% solve rate on the Eterna100 benchmark — comparable to published reinforcement-learning methods.
Research
RNA design · evolutionary computation · machine learning for structured biomolecules
DNA fluorescent aptamer design
Designed and computationally screened multivalent DNA Lettuce architectures, including a three-way-junction construct that positions two Lettuce aptamers on separate branches. Candidates were optimized and evaluated using oxDNA simulations, Boltzmann-based structural probabilities, and additional biophysical criteria. Eight selected designs are currently undergoing experimental validation.
RNA inverse folding and ribozyme design
Extending the cFold framework with additional folding oracles, a redesigned computational architecture, and support for pseudoknotted targets. The completed design pipeline has been applied to trans-acting hairpin and hammerhead ribozymes targeting an HTT-derived substrate, with experimental validation of catalytic activity underway. Complementary designs are being evaluated through SHAPE-based structural probing to directly assess whether the intended RNA structures are formed.