Kim Ihyeon

Computational Biology · Haeryong High School

Kim Ihyeon

I study where a protein ends up inside the cell — and how much of that a language model already knows.

I'm a student researcher at Haeryong High School working in computational biology. My current project applies semi-supervised learning to protein subcellular localization prediction, built on embeddings from the ESM-2 protein language model — aiming to make accurate predictions possible even when labeled data is scarce.

Input MKTAYIAKQRQISFVKSHFSRQLEERLGLIEVQ… a raw amino-acid sequence, no structure needed
Illustrative projection of ESM-2 embeddings, clustered by predicted compartment

01Research

Most localization datasets are small and expensive to label, while unlabeled protein sequences are abundant. My work explores whether ESM-2 embeddings already encode enough biological signal that a semi-supervised classifier can bridge that gap — turning sequence alone into a confident guess about where in the cell a protein belongs.

Nucleus Mitochondrion Cell membrane Cytoplasm Extracellular space

Model ESM-2 (Meta AI)  ·  Approach semi-supervised pseudo-labeling  ·  Task subcellular localization classification

02Recognition

Team Haeryong Cass

  • 2026 KYPT — Honorable Mention
  • Kwon Myeong-hoe Award (권명회상)

2026

03Get in touch