Search

The microbiome is a community of microorganisms living in our bodies and throughout the environment. The genomic data researchers can extract from microbiomes, known as metagenomic data, can be used to predict traits about a host or environment. By identifying microbiome biomarkers associated with disease or health, researchers can develop better therapeutics for microbiome-associated diseases. However, metagenomic data is commonly affected by technical variables unrelated to the phenotype of interest, such as sequencing protocol, which can make it difficult to predict phenotype and find biomarkers of disease. Here, we evaluate methods to remove background noise due to technical variables unrelated to the phenotype of interest, such as sequencing protocol, and thereby improving our ability to find accurate biomarkers of human disease. Also crucial in understanding host health is elucidating the sources of their microbiomes, as it allows researchers to understand the dynamics behind how microbial communities form and how they respond to changing environments. In this work, we introduce a method to use metagenomic variants obtained from hundreds of species in microbiome data to perform source tracking, which is a method of estimating colonization sources for a sample of interest. These analyses shed light on phenomena like the colonization of the early infant gut microbiome, or spatial patterns in the ocean microbiomes around the world. Lastly, we analyze metagenomic data to understand how genetic diversity changes along the human gut on the species, strain and gene level. In sum, this work leverages the genomic information contained in our microbiomes to find universal patterns in microbiomes, allowing us to better understand the relationship between microbiome and phenotypes, the colonization sources of microbiomes, and also the colonization dynamics on the species and strain level.

Background Illumina’s HumanMethylation450 arrays provide the most cost-effective means of high-throughput DNA methylation analysis. As with other types of microarray platforms, technical artifacts are a concern, including background fluorescence, dye-bias from the use of two color channels, bias caused by type I/II probe design, and batch effects. Several approaches and pipelines have been developed, either targeting a single issue or designed to address multiple biases through a combination of methods. We evaluate the effect of combining separate approaches to improve signal processing. Results In this study nine processing methods, including both within- and between- array methods, are applied and compared in four datasets. For technical replicates, we found both within- and between-array methods did a comparable job in reducing variance across replicates. For evaluating biological differences, within-array processing always improved differential DNA methylation signal detection over no processing, and always benefitted from performing background correction first. Combinations of within-array procedures were always among the best performing methods, with a slight advantage appearing for the between-array method Funnorm when batch effects explained more variation in the data than the methylation alterations between cases and controls. However, when this occurred, RUVm, a new batch correction method noticeably improved reproducibility of differential methylation results over any of the signal-processing methods alone. Conclusions The comparisons in our study provide valuable insights in preprocessing HumanMethylation450 BeadChip data. We found the within-array combination of Noob + BMIQ always improved signal sensitivity, and when combined with the RUVm batch-correction method, outperformed all other approaches in performing differential DNA methylation analysis. The effect of the data processing method, in any given data set, was a function of both the signal and noise. Electronic supplementary material The online version of this article (doi:10.1186/s12864-016-2819-7) contains supplementary material, which is available to authorized users.