MuziMùzǐ OuyangŌuyáng (Mia)

I design and test protein degraders, mostly bioPROTACs built on E3 ligase scaffolds like TRIM21, and use cancer drug resistance as a lens for understanding when and why they fail. My central finding came from comparing a tamoxifen-resistant breast cancer line (LCC2) to its drug-sensitive parent (MCF7): whole-genome sequencing turned up a mutation in the KRAS hypervariable region, a region with no previously known function, that nearly abolished the activity of a TRIM21-based KRAS degrader specifically in the resistant line. Beyond mechanism, I also build degraders as therapeutic candidates directly, including a tumor-specific bioPROTAC that co-degrades ERα and PI3K p85α to resensitize resistant breast cancer cells to tamoxifen, and, in an ongoing collaboration, a TRIM21-based bioPROTAC achieving dual-target degradation of oncogenic KRAS assemblies in pancreatic cancer.

The wet-lab puzzles above are usually what send me back to the computer. To explain why the KRAS mutation broke degrader activity, I ran 182-ns all-atom molecular dynamics simulations (AMBER, with explicit GppNHp and Mg²⁺) and showed that the mutation changes how KRAS clusters at the membrane, which the degrader’s activity turns out to depend on; swapping in a degrader architecture that doesn’t rely on clustering restored activity, confirming the mechanism computationally and experimentally at once. Separately, I taught myself a full de novo protein design pipeline, Rosetta Match for scaffold matching, ProteinMPNN for sequence design, and AlphaFold2 for structural validation, and used it to design new binders for small-molecule targets from scratch, open-sourcing the pipeline on GitHub. I see this computational side less as a separate skill set and more as the other half of the same question: once you know why a degrader fails, you can start predicting it.