|
Abstract:
|
The emergence of the microbiome as a central paradigm for many areas of biomedical research over the last 5 years has attracted numerous biologists and bioinformaticians to the field. During this time only a handful of biostatisticians have developed robust statistical methods for this data leaving the analytics to people with little or no formal training in statistical theory. My lab has developed Dirichlet-multinomial parametric methods for this data for formal hypothesis testing, sample size/power calculation, and parameter estimation which is needed to move microbiome research from exploratory to translational (bench-to-bedside) clinical research. In this presentation I will review the DM model and our prior work briefly, and focus on two problems we are currently addressing that extends the DM model to 1) Kullback-Leibler divergence to compare microbiome data sets, and 2) the importance of removing the concept of 'correlation' across microbe taxa which are guaranteed to be negative for this type of data.
|
