Skip to main content
Premium Trial:

Request an Annual Quote

Methods for More Accurate Microbiome Studies Could Impact Future Therapeutic Use


NEW YORK (GenomeWeb) – Researchers at the University of Minnesota took on the task of systematically analyzing biases in common methods for microbiome studies and developed a new set of "best practices" that could affect the therapeutic use of the microbiome.

In a study published online yesterday in Nature Biotechnology, the team documented four steps that labs can take to improve 16s ribosomal RNA amplicon-based marker gene surveys of microbial communities.

The group compared two methods — a commonly used protocol from the Earth Microbiome Project and a dual indexing protocol that was similar to one described in an Illumina technical note, according to the study.

There have been other investigations of issues in microbiome research methods, but the study was original in a few ways, according to the authors. "We used defined reference material, so that we had a ground truth to be able to compare to, and then we just systematically explored a large set of parameters and conditions," Daryl Gohl, first author of the study and a research and development lead at the University of Minnesota Genomics Center told GenomeWeb.

The project was part of a larger collaboration with the Mayo Clinic and was funded by a grant from the Minnesota Partnership for Biomedical Genomics. The expressed goal of that grant was to "solve problems in microbiome data generation," said Kenneth Beckman, a co-author of the study and the director of UMGC.

At the outset, it appeared from looking at the literature that "nobody had really gone through systematically and done proper controlled experiments — there were lots of papers that had done a little bit of poking at the problem, [but] we weren't satisfied ... so we decided we needed to do it ourselves," Beckman said.

Gohl noted that the process of preparing 16s amplicon libraries is seen as a deceptively simple thing. "It's just PCR — any first year grad student can go into the lab and do a PCR reaction — but that simplicity is deceptive, because it belies a huge amount of complexity in doing these multi-template reactions."

There are two classes of problem that arise in amplicon-based microbiome analyses, Gohl explained. "There's a quantitative bias that comes from amplification bias and different efficiencies of different templates, and there's also qualitative bias, which can cause the dropout of taxa due to primer mismatches."

The group found that quantitative biases could be improved by optimizing the polymerase enzyme, and limiting cycle numbers and template concentrations. And the enzyme of choice, the highly processive proofreading KAPA HiFi polymerase, also had the surprising benefit of improving the qualitative biases, the group discovered.

"We found that these enzymes were actually chewing back the primer sequences and allowing them to then match to templates that would have originally had mismatches," Gohl said.

The UMGC is now applying these best practices to about 30,000 samples per year, submitted by dozens of different investigators.  

How these biases could have affected previously published research that used other methods is unknown, but the group did attempt to model the quantitative effects by applying the error distributions it found in its analysis of a mock community to a large collection of published datasets.

The mock community, obtained from the Human Microbiome Consortium, consisted of 20 strains of bacteria in equal proportion. The study showed that the EMP method, which misestimated the abundance of three of the organisms by more than sevenfold, also led to complete failure to detect two of the organisms.

This dropout proved resistant to modeling. "We know the frequency of occurrence of these primer mismatches, but we don't know how often organisms that have those primer mismatches are present in actual samples, so ultimately we felt we couldn't effectively model that," Gohl said.

And the dropout is particularly problematic, Beckman noted, because there are situations where it might happen some times and not others.

"What we have right now is a system where there's a multitude of different methods used by different people, some of which will cause more dropout and some of which will cause less," he said. Comparing these data, researchers could mistake the occasional dropouts for signal.

"You might say, 'oh look, this species is now absent,' and you might actually assign biological meaning to that," said Beckman.

While irritating to researchers, the impact of quantitative and qualitative biases could be even more significant as microbiome analyses become clinical tools.

"The microbiome has been linked to all sorts of different diseases, so there is a very active development right now in pharma on being able to try to manipulate or modulate the microbiome in order to treat disease," Gohl said.

Gohl and Beckman noted there are a number of researchers pioneering fecal transplant, as well as companies — such as Seres, Rebiotix, and Assembly Bioscience — developing synthetic ecologies with the aim of displacing Clostridium difficile infections.

At the Personalized Medicine World Conference in Silicon Valley earlier this year, for example, firms discussed development of "curated therapeutic bacterial cocktails" which narrow down the healthy fecal microbiome to just the strains most likely to have a beneficial effect.

"Everyone is realizing that this is moving relatively rapidly ... toward application, so there's a big need to get the analysis up to speed," Beckman said, adding, "If people are going to be starting to treat and diagnose patients, you really can't have analytical methods that are flawed ... The question is, is anybody going to be happy having their microbiome manipulated based on anything less than the best possible analytical data?"

Indeed, as previously reported, the US Food and Drug Administration is now considering changing how it regulates fecal microbiota transplants for patients with recurrent C. diff.

However, advances may also be spurred on by the $121 million White House Microbiome Initiative launched earlier this year. When the project started, the National Institute of Standards and Technology said it planed to develop standards and measurement tools so discoveries made in different labs can be compared and replicated. Gohl and Beckman said that they will be attending an upcoming NIST workshop on microbiome measurement. "It will be interesting to get feedback," Beckman said.

The group is now actively working on the next generation of improvements to the protocol. "It's not like our paper is the final word; it's almost the first word," Beckman said.