Evaluating Correlation Measures for Inferring the Co-Occurrence Network in Microbial Communities

• Activity Number: 393
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Yuguang Ban*+ and Hongmei Jiang
• Companies: Northwestern University and Northwestern University
• Keywords: co-occurrence network ; correlation ; compositional data ; metagenomics

Abstract

The recent high-throughput sequencing has provided a high resolution of microbial organisms living in specific environments ranging from human gut to coal mining surface. Characterizing interactions among the organisms in a microbial community can give insights into how the microbial ecosystem is shaped under a specific condition. We study the co-occurrence pattern of microbial species across multiple subjects using their relative abundance information. However, it has been shown that bias can be introduced when correlations measured on compositional data. Yet, metagenomics data is known to be high dimensional and containing excess zero-count, and the compositionality bias has not been fully understood in this context. We investigate different distance and correlation measures using a simulation study. We also apply the measures on a real data. We show that inferring network structure from metagenomics data should be carried out carefully.