Oncolytic Viruses for Liver Cancer
Optogenetically Controlled Oncolytic Viruses for Hepatocellular Carcinoma Treatment
Oncolytic viruses (OVs) are a promising immunotherapeutic approach for liver cancer
treatment, but off-target toxicity and inflammation have limited their clinical
application. I designed a novel, optogenetically controlled OV system using the
light-sensitive photoreceptor PhoCl, which activates the OV only when exposed to UV
radiation. The construct achieved ~35% killing in aggressive liver cancer cell lines
over extended periods.
Cosmohedra: Physics-based Protein Prediction
Scalable prediction of symmetric protein complex structures
Accurately modeling protein structures that exceed ≈10,000 amino acids remains
difficult. Cosmohedra is a physics-based method to rapidly and accurately predict the
structure of arbitrarily large, symmetric protein complexes. Validated across 4 major
symmetry classes (icosahedral, tetrahedral, octahedral, and cyclic), the model
consistently achieves near-experimental levels of accuracy at more than 8 times the
limit of existing machine learning methods.
LLM-based Gene Delivery Development
Project EvoCapsid — Bio × ML Hackathon 2025
Abstract to come. Keep it to the same length as the others: the problem, what you
built, and the result.
COVID-19 Symptom Prediction
Machine learning drug discovery based on graph neural network and large language model
The COVID-19 pandemic presents an urgent need to understand the long-term health
implications associated with the virus, which include acute kidney injury, arrhythmia,
and stroke. I used a novel approach using a causal inference model to predict the
likelihood of post-COVID-19 symptoms, help identify patients at risk, and contribute
to the development of effective treatment strategies.
Modeling the Rate of Aging
Novel Epigenetic Clock based on DNA-methylation Trends and Aberrant RNA Splicing
Aging is an effective predictor for various cardiovascular, neurodegenerative, and
musculoskeletal diseases. To quantify aging based on variables including sex, disease
history, and life habits, I used a novel approach that combines DNA methylation rates
and aberrant RNA splicing to train a neural network that predicts aging with
near state-of-the-art accuracy (as of 2025).