Abundance or relative abundance - that is the question!

Speaker : Shyamal Peddada
Abstract:
Increasingly researchers are conducting microbiome studies to investigate differences in the composition of taxa in two or more ecosystems. Numerous statistical and computational methods have been proposed and used in the literature. However, in many cases it is not clear what statistical parameter is actually being tested in the null and alternative hypotheses. In this talk we describe two important statistical parameters of interest for comparing two or more ecosystems. We shall describe various methods used in the literature for testing hypotheses regarding these parameters. We shall specifically focus on a recently introduced methodology called ANCOM that takes into account the compositional structure of microbiome data, and demonstrate, using simulation studies, that among the available methods known to us, this is the only method that has a good control of false discovery rate (FDR) while maintaining high power. In addition to compositional structure in microbiome data, these data contain excess amounts of zeros. In some cases, as much as 90% of the OTU table may contain zeros. In this talk we describe three potential sources of zeros and briefly describe a simple methodology to identify the potential source of zero and how to handle them in the analysis.