Hyungwon CHOI 
    Main Page   Publications              
a filler
  Hyungwon CHOI  
  Lab Location: 6-08B

tel: 65869780

Email: hwchoi@imcb.a-star.edu.sg
a a a
a a a
  Key Publications  

Z. Cheng#, G.S. Teo#, S. Krueger, T.M. Rock, H.W.L. Koh, H. Choi*, C. Vogel*.
Differential dynamics of the mammalian mRNA and protein expression response to misfolding stress.
Mol. Syst. Biol. In press (2016).

G.S. Teo, S. Kim, Ben Collins, A.C. Gingras, A.I. Nesvizhskii, H. Choi.
mapDIA: a new platform for data normalization and analysis for label-free quantitative proteomics data.
J. Proteomics. 129:108-20 (2015)

H. Koh, H. Swa, D. Fermin, S.G. Ler, J. Gunaratne, H. Choi.
EBprot: differential expression analysis in labeling-based quantitative proteomics experiments.
Proteomics. 15(15):2580-91 (2015)

G.S. Teo, C. Vogel, D. Ghosh, S. Kim, H. Choi.
PECA: a novel statistical tool for deconvoluting time-dependent gene expression regulation.
J. Proteome Res. 13(1):29-37. (2014)

D. Fermin, S. Walmsley, A-.C. Gingras, H. Choi*, A. Nesvizhskii*.
Luciphor: algorithm for phosphorylation site localization with false localization rate estimation using modified target-decoy approach,
Mol. Cell. Proteomics 12: 3409-3419 (2013)

H. Choi, D. Fermin, A. Nesvizhskii, D. Ghosh, Z.S. Qin.
Sparsely correlated hidden Markov models with application to genome-wide location studies.
Bioinformatics 29(5):533-41 (2013)

H. Choi, B. Larsen, Z.-Y. Lin, A. Breitkreutz, D. Mellacheruvu, D.Fermin, Z.S. Qin, M. Tyers, A.-C. Gingras, A. Nesvizhskii.
SAINT: probabilistic scoring of AP-MS data.
Nat. Methods 8:70-73 (2011)




a a a
  Hyungwon CHOI

Hyungwon Choi obtained Ph.D. in biostatistics from University of Michigan (2009) and worked as post-doctoral research fellow in the University of Michigan Medical School (2009-2010). During his time in Michigan, he developed a wide range of computational and statistical algorithms for analyzing various types of high-throughput molecular datasets. In 2011 he joined National University of Singapore as assistant professor in biostatistics, where he has developed mass spectrometry data processing methods and pathway-oriented statistical models for proteomics and metabolomics data in clinical and basic science applications. He has also held Adjunct PI position at IMCB since February 2014.

  Summary of Research

My research interest is in developing computational methods for the analysis of complex molecular data sets from experimental platforms such as next generation sequencing (NGS) and mass spectrometry (MS). The main themes of my research are (i) reliable extraction of qualitative and quantitative information from noisy experimental data, and (ii) integration of multiple data sources for generating specific biological hypotheses from such large-scale -omics data sets. Our current research projects include:

Computational Proteomics and Metabolomics

  • Development of informatics tools for preprocessing and quantitation of proteins and metabolites
  • Spectral assay library development for small molecules and proteins
  • Integrative analysis of quantitative proteomics data and protein structure information

Statistical Methods for Systems Biology

  • Network-based statistical methods for identifying differentially regulated molecules
  • Integrative analysis of multi-omics data for elucidating the mechanism of dynamic gene expression regulation
  • Visualization of multi-omics data

Applications to Translational Research

  • Proteogenomic analysis of large-scale multi-omics datasets in human cancers
  • Data mining of molecular datasets for host-pathogen interaction in immunology applications
  • Metabolomics analysis in infectious diseases and nutrition epidemiology

( Click to view larger image )

Figure legend:
A schematic view of our multi-omics data integration strategy, with illustration from the PI3K-AKT-mTOR signaling pathway in ER+/PR+ (luminal), HER2 enriched (HER2E), and basal-like tumors in the Cancer Genome Atlas data (invasive breast cancer). A collection of curated biochemical pathways is compressed into a mosaic of simpler undirected graphs, and network motifs of moderate sizes (e.g. 3-5) are captured. Quantitative measurements from each –omics platform are translated into (putative) stoichiometric ratios based on various biological networks, and these network-derived quantitative data are contrasted among different cancer subtypes. An efficient statistical method incorporating the topology information of the underlying biological network, called Markov random field modeling, is employed to detect the most promising sub-networks that differentiate different subtypes. This approach improves the chance of finding biologically relevant, therapeutically amenable differential markers compared to the conventional alternative approach based on an amalgam of univariate statistical hypothesis testing.  Shown in the figure is an example of PI3KCA to PDK1 ratio, which is low in basal-like and HER2E subtypes and high in luminal subtypes, which would not have been statistically significant if individual proteins were tested for differential expression.