Toward More Reproducible, Reusable Stable Isotope Probing Experiments

As SIP experiments shed more light on microbial communities than ever before — especially with the prospect of meta-analyses boosted by AI and Machine Learning — JGI researchers and collaborators work to create FAIR datasets to maximize returns on these advances.

Over the last two decades, Stable Isotope Probing (SIP) has opened vast windows into specific activities catalyzed by microbial communities. By pairing RNA or DNA sequencing with tracings of stable, heavy isotopes, SIP allows researchers to go beyond understanding which microbes are present in a sample, to generate data about which organisms are active, and which microbes are eating what,  in the environment. 

This distinction — between presence and participation — is especially important in microbial ecosystems. These communities harbor tens of thousands of different organisms who all consume different compounds. Microbial species will vary in activity as well; at any given time, many species in a community may be in ‘low power mode’, where they do very little. 

SIP experiments can show which species actively munch on a specific chemical compound. By feeding a microbial community a modified version of a specific chemical, where ordinary Carbon-12 atoms are swapped for the heavier, non-radioactive isotope Carbon-13, researchers can then safely follow the path of that modified chemical in any environment. Organisms who process the modified compound will have heavier DNA or RNA, and SIP techniques leverage this to untangle which microbes ate the compound and how much they ate. 

These analyses provide granular insight into microbial functions, traits and competition, illuminating the inner workings of an environment. For example, SIP experiments can show how soil microbes cycle nitrogen and carbon, or how freshwater ecosystems process methane and organic matter. 

However, to get all of the information that SIP offers, researchers must take on a fairly time-intensive process. The isotope labeling incubations can take hours to months, followed by weeks of tricky, laborious lab work to analyze samples from one experiment.

Starting a  few years ago, researchers can request quantitative SIP analysis as a capability in proposal calls to the U.S. Department of Energy (DOE) Joint Genome Institute (JGI), a DOE Office of Science User Facility located at Lawrence Berkeley National Laboratory (Berkeley Lab). “By offering Stable Isotope Probing as a service at JGI, we're trying to bring in a lot more people who want to do it but just otherwise couldn't or wouldn't,” said Rex Malmstrom, leader of the Microscale Applications group at the JGI, whose team uses a semi-automated, standardized process to carry out the onerous SIP sample processing.

Generating these datasets opens up the possibility of new insights across ecosystems and research teams. Even better, creating comparable data with useful and accessible metadata multiplies these possibilities, and maximizes returns from the demanding nature of SIP experiments. 

Especially as machine learning and AI promise new ways of harnessing large datasets, JGI researchers and users are looking for ways of making these experiments more repeatable and reproducible. Essentially, the aim is to ensure these experiments create data aligned with FAIR principles (Findability, Accessibility, Interoperability, and Reusability) for data reuse. 

In the last year, JGI’s user community has helped propel multiple efforts along these lines, all with the goals of better data reporting, improved collaboration, and more comparable analysis.

A Data Standard for SIP 

Recently, Purdue researcher and JGI user Roli Wilhelm set out to do a SIP-fueled meta-analysis with a member of his lab, Abigayle Simpson. They found that although 20 years of SIP experiments have generated plenty of insights, many studies don’t report the information needed to reuse and compare data.

“So we thought, let's go on a pilgrimage. Let's make a difference, from a different direction — and that's helping people do a better job curating the data,” Wilhelm said. Informed by other community-driven standards from the Genomic Standards Consortium, Wilhelm and Simpson assembled a team of researchers working in this SIP space to formalize metadata reporting and data labeling for these experiments. 

This group included Elisha Wood-Charlson, User Engagement Lead for the DOE Systems Biology Knowledgebase (KBase), as well as researchers from Pacific Northwest National Laboratory, Northern Arizona University, Argonne National Laboratory and the JGI. 

With input from this group, as well as feedback from researchers around the world, the team have now published a Minimum Information for any Stable Isotope Probing Sequence (MISIP). They differentiate between required information and terms, such as isotopes involved in an experiment, and recommended information, such as additional substrates. 

The idea is to improve the information produced in the next few decades of SIP experiments. “I’m looking forward to seeing that come through, predicated on these data standardization efforts — that as the data gets churned out, we start to really see the utility of them broaden,” Wilhelm said. 

Sharing Ideas to Improve Experiments

In a separate effort, JGI users and collaborators gathered for a SIP-focused workshop at the JGI in the fall of 2024. “ We wanted to get into the nitty gritty with the people doing SIP experiments — how do we do this better? How do we improve data accuracy and make results from different SIP studies easier to compare?” said Malmstrom, who worked with JGI Metagenome Program head Emiley Eloe-Fadrosh to host researchers in Berkeley. 

Organizing this workshop, too, was collaborative. Bruce Hungate and Egbert Schwartz of Northern Arizona University, and longtime JGI user Jennifer Pett-Ridge of Lawrence Livermore National Lab also played key roles in creating sessions where attendees could focus on improving data collection, analysis and reporting. 

Thirty-four attendees came together for the workshop, including Wilhelm. “ This workshop was really nice because people were given structure, but it was all about sharing ideas. All breakout sessions, just enough context to get people thinking and tackling some of the issues. It was really productive,” he said. 

Data reporting standards, like Wilhelm’s MISIP framework, were interesting to lots of workshop attendees. As a group, these researchers all aim to build awareness around the value of metadata reporting. “ If you collect and provide this metadata, you are now enabling these much larger comparisons down the line,” Malmstrom said.

In addition to improving data reusability, the workshop gave researchers a chance to come up with ways to measure the reproducibility of results between groups. 

Schwartz and Malmstrom came away from the workshop with a concrete experimental plan for a ring test, or a survey. Together, they will set up a broad experiment to compare results between labs. Their labs, as well as Pett-Ridge’s group, will assemble input materials, with eight or nine different teams then performing analyses that they will all compare. Thanks to robust internal analytical standards, the JGI is well-positioned to help with this kind of intercomparison. 

After a few decades, SIP techniques are well-primed for these kinds of considerations, although reusability and reproducibility are still growing priorities for researchers in this area. “Surprisingly, this was the first workshop of its kind on SIP metagenomics,” Malmstrom said. 

Increasingly, efforts like these will help grow and improve high-quality datasets, enabling global insights from modeling, machine learning, and subsequent experiments.