April 13, 2022
What is Differential Abundance (DA)?
The human microbiome is composed of a wide variety of microbial species, working in harmony with each other and with our cells to regulate normal functions like digestion. However, if the delicate balance of the relative abundance of different microbial species is disturbed, the side effects can be extremely unpleasant, or potentially even deadly.
Thus, it is vitally important to be able to characterize the composition of microbial communities, both by identifying the different species present and their relative abundance to each other. This second marker, called differential abundance or DA, answers a seemingly simple question: which species differ significantly in their relative abundance between samples?
Problems With Current Microbiome DA Methods
Interestingly, though there are a large number of tools available for DA testing, there is no consensus on the best practices for this kind of analysis. The raw data for DA analysis is DNA sequencing data of a characteristic and abundantly expressed marker gene, typically the 16S ribosomal RNA subunit gene. Data pre-processing can include correcting for read depth and filtering out rare species, but different computational tools may use neither of these approaches, one of them, or both.
As for the actual computational tools for DA analysis, their performance varies based on the distribution type of the data being analyzed, and they also vary dramatically in their statistical power. When evaluated using simulated datasets with no expected differences, many of these tools produced extremely high false discovery rates of significant differences in abundance.
An Independent Analysis of DA Methods
With all of the inconsistencies in available tools for such critical analysis for microbiome studies, Nearing et. al. undertook an independent evaluation of 14 existing DA tools and methods using 38 datasets, published in Nature Communications.
Their findings confirmed the variation between these tools, with the tools identifying dramatically different numbers of significant differences between groups, and found that the variation depended most on how the data was pre-processed.
Two tools, ALDEx2 and ANCOM-II, produced the most consistent results and carried the most in common across results from different tools, but the authors concluded that even these tools are not sufficiently accurate to extrapolate the findings to robust biological conclusions. The authors recommended a consensus approach be taken for microbiome analysis, using multiple DA methods and extracting common results.
The applications of these findings are significant. Existing microbiome DA analyses need to be critically examined through the lens of the accuracy of the data processing and bioinformatic tools used for analysis to determine if the biological conclusions they yielded hold up. The human microbiome affects everything from the immune system to the digestive system, so it is incredibly important that studies in this area continue to advance, and that the computational tools and methods used in the field advance as well.
Outsourcing Bioinformatic Analyses
This study highlights the importance of quality data pre-processing and having the appropriate computational tool(s) at your disposal. If the tools you need aren’t performing, or don’t exist, consider reaching out to a bioinformatics as a service (BaaS) provider like Bridge Informatics. Skilled computational scientists with the bench experience and biological knowledge to understand your project are what set Bridge Informatics apart from other BaaS providers. Book a free discovery call with us today to see if we can meet your data analysis and pipeline development needs.
Jane Cook, Journalist & Content Writer, Bridge Informatics
Jane is a Content Writer at Bridge Informatics, a professional services firm that helps biotech customers implement advanced techniques in management and analysis of genomic data. Bridge Informatics focuses on data mining, machine learning, and various bioinformatic techniques to discover biomarkers and companion diagnostics. If you’re interested in reaching out, please email daniel.dacey@old.bridgeinformatics.com or dan.ryder@old.bridgeinformatics.com.
Sources:
https://www.nature.com/articles/s41467-022-28034-z
