• Developed new statistical techniques and applied machine learning methods to find mutations that cause rare genetic disorders.• Performed perturbation experiments on NLP models to gain a better understanding of their inner workings.Selected PublicationsSeplyarskiy V*, Koch E*, Lee D* et al. 2023 “A mutation rate model at the basepair resolution identifies the mutagenic effect of polymerase III transcription” Nature Genetics* indicates equal contribution• Developed the mutation rate model that outperforms all previous methods by using logistic regression based on various genomic and epigenomic features.• Modified Naglkerke’s R2 to compare the performance of mutation rate models in an interpretable way.Lee, D, and Heimersheim S. 2024. “Investigating Sensitive Directions in GPT-2: An Improved Baseline and Comparative Analysis of SAEs.” arXiv• Accepted to NeurIPS 2024 ATTRIB WorkshopLiu A, Lee D, Jalihal A, Hanage W, Springer M. 2023 “Quantitatively assessing early detection strategies for mitigating COVID-19 and future pandemics” Nature CommsConnally N, Nareen S, Lee D, Sunyaev S, et al. 2022 “The missing link between genetic association and regulatory function” eLifeManuscriptLee D*, Koch E* et al. 2024 “multiSFS: using deviation of multinomial SFS to detect selection across the genome”• Developed a new statistic using tail of a distribution to estimate negative selection from genomic sequencing data. This method outperforms all previous genome-wide statistics.Other Projects• Developed constrained Bayesian multinomial regression in Pyro to infer negative selection from various genomic and epigenomic features.• Used pretrained NLP models in Hugging Face to classify variants of uncertain significance as pathogenic or benign by training on ClinVar text data.

Skill up on AI safety, covering topics such as mechanistic interpretability, eliciting latent knowledge, learning from human feedback, goal misgeneralization, and scalable oversight.

• Applied multiple sequence alignment among GPCR families, genetic data, and protein features to find promising drug target.

Studied the biophysical interaction among breast cancer cells using the Cellular Potts model in C++.PublicationsTran-Ba K-H, Lee D, Zhu J, Paeng K, Kaufman LJ "Simultaneous Confocal Microscopy and Rheology Probes the Structural and Mechanical Evolution of Collagen I through the Sol-Gel Transition” Biophysical Journal, 2017• Analyzed the fractal dimension of a collagen network to find evidence for Sol-Gel transition.