M U Kim

The paper evaluates which layers of genomic foundation models are most critical for variant classification, finding that representative and load-bearing layers differ in their roles and positions within the model.

This paper uses holotomography to analyze 3D biophysical changes in mouse oocytes during post-ovulatory aging, showing that non-morphological features better distinguish aged from fresh oocytes.

DeepHOPE uses deep learning and 3D imaging to assess pluripotency in hPSCs, detecting structural changes before molecular loss of pluripotency.