2016 AIChE Annual Meeting
(228dr) Lipidomic and Transcriptomic Biomarkers for Diagnosis of Nonalcoholic Fatty Liver Diseases
Authors
Using lipidomic and transcriptomic data from about 90 human liver-tissue samples, we developed a LDA-based classifier. We found that about 75-85 variables (genes or lipids) are needed to accurately classify among all the four classes/conditions (normal, steatosis, NASH and cirrhosis) whereas about 20 genes or lipids were enough to classify between steatosis and NASH samples. These findings are in contrast to our initial expectation to be able to find a small number of biomarkers, for example, 5 or less genes or lipids, for accurate classification. Two likely reasons for the need of a larger number of biomarkers for complete diagnosis are: (1) relatively much larger variability in the measurements in tissue-level samples as compared to that in cell culture, and (2) the complex and systems-level nature of the disease. Towards the latter possibility, we carried out KEGG pathway and GO-term enrichment of the genes differentially regulated among the four classes or between steatosis and NASH. We found that pathways such as focal adhesion and ECM-receptor interaction were enriched in NASH vs. steatosis comparison suggesting increased inflammation during the progression from steatosis to NASH. Fibrosis pathway and collagen-related genes were enriched in the overall comparison, indicating their role during the late stage NASH and cirrhosis. Using PLINK, we also analyzed single nucleotide polymorphism (SNP) data from the samples. The well-known SNP marker in PNPLA3, rs738409, showed a strong statistical significance with a p-value of 1.7E-6 between normal and cirrhosis samples, but not between NASH and steatosis samples. We are continuing further integrated analysis of the gene-expression and SNP data. Overall, we conclude that progression of NAFLD involves many genes, lipids and pathways and only a systems-based classification and analysis approach will lead to accurate/reproducible diagnosis and mechanistic understanding of NAFLD.
Acknowledgements:This study were primarily supported by the National Institutes of Health (NIH) glue grant GM U54069338 to the LIPID MAPS consortium. We would also like to acknowledge the National Science Foundation (NSF) collaborative grant STC-0939370, and NIH R01 grants HL087375-02, HL106579 and HL108735.
References:
Gorden, D. L., D. S. Myers, P. T. Ivanova, E. Fahy, M. R. Maurya, S. Gupta, J. Min, N. J. Spann, J. G. McDonald, S. L. Kelly, J. Duan, M. C. Sullards, T. J. Leiker, R. M. Barkley, O. Quehenberger, A. M. Armando, S. B. Milne, T. P. Mathews, M. D. Armstrong, C. Li, W. V. Melvin, R. H. Clements, M. K. Washington, A. M. Mendonsa, J. L. Witztum, Z. Guan, C. K. Glass, R. C. Murphy, E. A. Dennis, A. H. Merrill, Jr., D. W. Russell, S. Subramaniam and H. A. Brown, â??Biomarkers of NAFLD Progression: A Lipidomics Approach to an Epidemicâ?, Journal of Lipid Research, 56(3), 722-736, 2015.
Corresponding author: Shankar Subramaniam, Department of Bioengineering, University of California San Diego, 9500 Gilman Drive, La Jolla, CA, 92093-0412, Phone: (858) 822-0986, E-mail: shankar@ucsd.edu.