Hoon (Hyunghoon) Cho
Ph.D. in Computer Science, Massachusetts Institute of Technology, 09/2013 - Current
Advisor: Prof. Bonnie Berger
M.S. in Computer Science, Stanford University, 03/2012 - 06/2013
B.S. in Computer Science with Honors, Stanford University, 09/2009 - 06/2013
Honors Thesis: Unraveling the genetics of human diseases by integrating patterns for epistasis detection. [ PDF ]
Advisor: Prof. Daphne Koller
Korea Science Academy of KAIST, 03/2006 - 12/2008
Mashup: Compact Integration of Multi-Network Topology for Functional Analysis of Genes
Hyunghoon Cho, Bonnie Berger, Jian Peng
Under review, 2016
Reconstructing Causal Biological Networks through Active Learning
Hyunghoon Cho, Bonnie Berger, Jian Peng
PLoS ONE 11(3), 2016 [ Link ]
Exploiting Ontology Graph for Predicting Sparsely Annotated Gene Function
Sheng Wang*, Hyunghoon Cho*, ChengXiang Zhai, Bonnie Berger, Jian Peng
ISMB/ECCB, 2015. Bioinformatics 31(12), 2015: i357-i364. [ Link ]
Diffusion Component Analysis: Unraveling Functional Topology in Biological Networks
Hyunghoon Cho, Bonnie Berger, and Jian Peng
Accepted for presentation at RECOMB 2015.
High-Resolution Transcriptome Analysis with Long-Read RNA Sequencing
Hyunghoon Cho, Joe Davis, Kevin S. Smith, Alexis Battle, and Stephen B. Montgomery
PLoS ONE 9(9), 2014 [ Link ]
Bayesian clustering of mixed membership modules across multiple networks with Gerald Quon and Prof. Manolis Kellis. 09/2013 - present.
I am currently developing a Bayesian hierarchical model for identifying gene modules that are consistent across multiple networks within a population of networks. This model can be applied to transcriptional and regulatory networks that are constructed from multiple cell types or individuals to uncover meaningful natural variations in network structure.
Transcriptome analysis with long-read RNA sequencing with Joe Davis, Alexis Battle, and Prof. Stephen Montgomery. 09/2012 - 06/2013.
Ongoing improvements to next generation sequencing technologies are leading to longer sequencing read lengths, but a thorough understanding of the impact of longer reads on RNA sequencing analyses is lacking. To address this issue, we generated and compared two RNA sequencing datasets of differing read lengths (2x75 bp and 2x262 bp) and investigated the impact of read length on various aspects of analysis, including the performance of currently available read-mapping tools, gene and transcript quantification, and detection of allele-specific expression patterns.
Patterns of epistasis and guided adaptive interaction testing with Alexis Battle and Prof. Daphne Koller. 04/2011 - 09/2012.
One of the biggest challenges in modeling the genetics of complex human diseases is in discovering important interactions (non-additive effects) of multiple genetic variants such as single nucleotide polymorphisms. Because of the combinatorial explosion in the number of interaction hypotheses to consider, most naive approaches suffer from computational burden and the loss of the ability to detect small, but important, interactions due to multiple hypothesis testing correction. We developed a method that incorporates several important patterns of gene-gene interactions and learns how to prioritize candidate interactions. This resulted in a discovery of hundreds of significant interactions in several datasets that would not have been statistically significant if we took the naive approach.
Honors research project. [ Thesis ]
Submitted to Genome Research (under revision).
Using the Deformable Part Model with autoencoded feature descriptors for object detection with David Wu, Adam Coates and Prof. Andrew Ng. 09/2010 - 01/2011.
The Deformable Part Model (DPM) is widely regarded to be one of the state-of-the-art object detection algorithms. We considered substituting the human-engineered Histogram of Oriented Gradients descriptors used by DPM for features learned in an unsupervised fashion by a single-layered, sparse autoencoder. The results indicated that the unsupervised feature learning-backed DPM achieves a comparable performance as the original HOG-based system but does not show significant improvement.
Frequency dependence of radiation patterns of earthquakes on rough faults with Prof. Eric Dunham. 06/2010 - 12/2010.
The focus of this project was to understand far-field ground motion from simulations of rupture propagation on rough faults. The observations suggested that the directivity effect (which states that the ground motion in the forward direction of the rupture propagation tends to be larger than that in other directions) is generally only present at frequencies less than about 1 Hz, and that at higher frequencies, the directivity effect vanishes and the far-field radiation pattern changes from the usual double-couple pattern to an isotropic one.
Teaching Assistant for CS 228: Probabilistic Graphical Models taught by Prof. Daphne Koller, Stanford University, 03/2013 - 06/2013.
Section Leader for CS 106A: Programming Methodology and 106X: Programming Abstractions, Stanford University, 01/2011 - 12/2011.
Research Intern, Microsoft Research New England, Summer 2014.
Business Development Intern, Palantir Technologies, Summer 2012.
MIT EECS Great Educators Fellowship, 2013.
Kwanjeong Educational Foundation Scholarship for Graduate Studies, 2013.
Frederick Emmons Terman Engineering Scholastic Award, 2013
Stanford Tau Beta Pi Engineering Honor Society, 2012 - present.
President's Award for Academic Excellence in the Freshman Year, 2010.
Seventh Place in ACM ICPC Pacific Northwest Regional, 2010.
Kwanjeong Educational Foundation Scholarship for Undergraduate Studies, 2009.
Second Award in Intel ISEF, 2008.