Advancing Bioinformatic Technologies for Lipidomic Discovery: Lipid Identification, Pathway Analysis, Network Analysis

点击翻译按钮获取中文摘要

Lipidomics: Lipidomics is the simultaneous measurement of thousands of individual lipid species realized through a combination of lipid biochemistry, mass spectrometry (MS), and bioinformatics. Spatio-temporal maps of lipid compositions represents a robust "readout" of an organism's metabolic response to genetic and environmental change as well as a unique fingerprint of their metabolic identity. Despite this promise, we lack the necessary bioinformatic tools to connect data-driven lipidomics with network analyisis. Mining datasets remains labour-intensive and time-consuming. There is little standardization between laboratories and, to our knowledge, no data processing pipelines capable of processing both untargeted and targeted lipidomic approaches. At least four bioinformatic challenges must be addressed: (1) Data processing and lipid identification; (2) Statistical analysis; (3) Pathway and network analysis; and (5) Predictive systems modelling in biophysical context. Objectives: With 5 years of NSERC support, my team developed new bioinformatic tools that faciliatate identification and quantification of lipid species in different samples from the same matrix and more time-consuming absolute quantification of lipid species in different matrices across MS platforms. These tools (RTStar, LIT, LITL, and VaLID) focus on problems of Data Processing and Lipid Identification. Here, I will expand our bioinformatic research program to focus on problems associated with Statistical Analysis and Pathway and Network Analysis. I hypothesize that current limitations on identifying and interpreting lipidomic data can be overcome with a combination of machine learning, probabilistic modeling, automatic network construction, visualization and causal network analysis. My team's short term objectives (1-3 years) are to expand our lipid identification tools by incorporating probabilistic modeling, new Bayesian statistical methods and automatic pathway visualization to more rapidly and efficiently assign lipid identities and map consequences of changes in lipid abundances. Our medium term objectives (3-5 years) are to build multi-scale, causal network models that enable interpretation of lipidomic datasets tested and validated using population-based data generated in our laboratories and others. Our long term objectives (>5 years) will be to develop new predictive modeling approaches that incorporate genomic and proteomic data into causal lipidomic networks combining "omic" approaches to map lipid metabolism. Training: This proposal is supported by an NSERC CREATE Training Program Metabolomics Advanced Training and International Exchange (MATRIX) awarded Sept 2018 (PI: Dr Mary-Ellen Harper; I am the co-Director of MATRIX-CREATE). MATRIX-CREATE will ensure trainees have extensive expert exposure to all aspects of metabolomics and lipidomics, with particular focus on bioinformatics.