(541a) Statistics and Pattern Discovery: Identification of Metabolomic Biomarkers for End Stage Renal Disease | AIChE

(541a) Statistics and Pattern Discovery: Identification of Metabolomic Biomarkers for End Stage Renal Disease

Authors 

Stephanopoulos, G. - Presenter, Massachusetts Institute of Technology
Tong, L. - Presenter, Massachusetts Institute of Technology
Thadhani, R. - Presenter, Massachusetts General Hospital


Metabolomics has emerged in recent years as the field dealing with the measurement of the entirety of small metabolites in cells. While complementing information obtained from transcriptional and proteomic analysis, metabolomics also offers additional insights into the kinetics of internal metabolism, higher sensitivity and resolution, and response at shorter time scales relatively to proteins or transcripts. As such, besides their obvious importance to understanding basic biological mechanisms, metabolite measurements are also important as biomarkers in disease diagnosis and treatment. Advanced and specialized methods are needed, however, to unveil discriminating metabolites and their patterns that separate distinct classes of patients with respect to health status and outcome of disease. Such methods make use of contributions in statistics made by Professor Ramkrishna.

Prior research in our laboratories has solved two important problems in high throughput metabolomic analysis, namely: (a) the limited availability of metabolite standards for MS peak identification, and, (b) the lack of standardized procedures for very high measurement reproducibility and differential metabolite change detection, of the quality required for high fidelity biomarker identification. A new tool, SpectConnect, has been made available in the web for tracking and cataloguing otherwise unidentifiable but conserved metabolites across sample replicates without use of reference spectra. In addition, we have successfully tested these methods in the analysis of 120 plasma samples from Early Stage Renal Disease (ESRD) patients thus identifying distinct biomarkers differentiating patient survival after 90 days of hemodialysis. We will review these developments in this talk and illustrate the prospect of metabolomics in identifying powerful biomarkers with superior ROC characteristics relatively to other clinical and epidemiological data.