Integrative OMICs Analysis in Quantifying Tissue Specificity

• Activity Number: 618 - Machine Learning for Big Data
• Type: Contributed
• Date/Time: Thursday, August 1, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Meng Wang* and Lihua Jiang and Hua Tang and Michael Snyder
• Companies: Stanford University and Stanford University and Stanford University and Stanford University
• Keywords: robust estimation; data adaptive; tissue specificity; OMICs data; integrative analysis

Abstract

To accurately detect and quantify gene tissue specificity (TS) is important to help researchers to further expand the understanding of tissues functions at the molecular level, and provides significant insights into disease mechanisms and tissue-specific therapeutic targets. High-throughput OMICs data improves the study of TS. The GTEx consortium has established genome and transcriptome databases from multiple tissues and the eGTEx Phase I study has generated proteomics data from 32 tissues. Integrating RNA and protein expressions to understand tissue specificity is very meaningful but also challenging due to different OMICs from different platforms and different noise levels. Our contributions are in threefold: (1) in the preprocessing step, we develop a new robust normalization method to adapt tissue sample heterogeneities, (2) in the step of quantifying TS, we build data-adaptive robust z-scores to make TS comparable across OMICs, (3) in the integration step, we study protein and RNA concordance and discordance in TS from 2D joint robust fitting. Our integrative analysis is based on robust estimation to quantify TS to better understand protein and RNA regulations across tissues.