Seyoung Kim

Contributions to Public Health

• To understand the genetic basis of phenotypic variability, most existing methods for eQTL mapping and GWAS examined a single phenotype and a single genetic variant at a time. Instead, taking systems-genetics approach, my lab developed statistical machine learning methods for learning a gene network and eQTLs that perturb this network simultaneously, given eQTL datasets, and further extended this to integrate multiple omics data with a highly-scalable learning algorithm.
  • Zhang L and Kim S. Learning gene networks under SNP perturbations using eQTL datasets. PLoS Comput Biol., 10(2):e1003420, 2014. PubMed PMID: 24586125; PubMed Central PMCID: PMC3937098.
  • McCarter C, Howrylak J, and Kim S. Learning gene networks underlying clinical phenotypes using SNP  Perturbation. PLoS Computational Biology, 16(10):e1007940, 2020.
• Allele-specific expression is essential for understanding gene regulation in diploid organisms. My lab developed computational tools for highly accurate and efficient allele-specific expression quantification via a light-weight extension of the popular software kallisto, and for learning gene regulatory networks and cis-acting/trans-acting eQTLs from eQTL data with allele-specific expression.
  • Adduri A. and Kim S. Ornaments for efficient allele-specific expression estimation with bias correction. The American Journal of Human Genetics (accepted).
  • Yoon J and Kim S. Learning gene networks under SNP perturbation using SNP and allele-specific expression data, BioRxiv 563661 [Preprint]. Oct 24, 2023. Available from: https://doi.org/10.1101/2023.10.23.563661.
• For longitudinal measurements of high-dimensional correlated features such as spatio-temporal data for COVID case counts, it is essential to model both the structure over the multivariate features and their evolution over time jointly with a highly scalable learning algorithm. My lab developed flexible nonparametric Bayesian methods based on Gaussian processes for multivariate-output regression with time as covariates.
  • Yoon JH, Jeong D, and Kim S. Doubly mixed-effects Gaussian process regression. In Proceedings of the Conference on Artificial Intelligence and Statistics (AISTATS), pages 6893–6908. PMLR, 2022.
  • Jeong D and Kim S. Factorial stochastic differential equation for multi-output Gaussian process regression. In Proceedings of the Conference on Artificial Intelligence and Statistics (AISTATS), pages 9755–9772. PMLR, 2023.
• In many real-world data, complex dependencies are present both among samples and among features, such as expression measurements of tens of thousands of genes collected for individuals in a pedigree. My lab developed statistical methods for jointly modeling dependencies among samples and features.
  • Yoon JH and Kim S. EiGLasso for scalable sparse Kronecker-sum inverse covariance estimation. Journal of Machine Learning Research, 23(110):1–39, 2022.

Complete List of Published Work in MyBibliography