Billion-match CRISPR spacer database links microbes to their viruses
Researchers at the US Department of Energy Joint Genome Institute (JGI) have built a bioinformatics toolkit, SpacerExtractor, and a matching database of CRISPR spacers drawn from roughly 400,000 public metagenomes. The resource holds more than 100 times as much CRISPR data as earlier databases, and its spacers produced over a billion matches to viral sequences. The work, led by Simon Roux and colleagues, is reported in Nature Microbiology.
The host problem in virus research
Most viruses of bacteria and archaea are known only from sequencing data, because they have never been grown in the laboratory. For these viruses, the microbes they infect are usually unknown. That gap limits efforts to understand how viruses shape microbial evolution and metabolism, whether by killing their hosts or, at times, by temporarily supplying them with extra capabilities.
CRISPR-Cas systems offer a way round the problem. These immune defences occur in about 40% of bacterial and 85% of archaeal genomes. They store short DNA fragments, called spacers, taken from mobile genetic elements such as viruses, and so act as a record of past encounters. A spacer that matches a viral genome points to a microbe that has defended against that virus, and therefore to a likely host.
Scale and infrastructure
The JGI team developed SpacerExtractor to pull spacers out of the metagenome collection. Moving and processing data on this scale relied on the Energy Sciences Network (ESnet), a Berkeley Lab user facility that provides high-speed data transfer. Researchers can use the database in two directions: to check the probable hosts of a newly found virus, or to look up the viruses linked to a microbe they study.
What the matches show
The analysis found that viral infections often lead to sustained CRISPR targeting, and that spacers can cover the whole viral genome. Some regions of viral genomes cannot easily mutate to escape targeting. Spacers aimed at those regions are probably more effective and, in time, more common in microbial populations.
About 10% of targeting did not fit this pattern and was not linked to any change in the viral genome. The authors suggest this may reflect a different mode of virus-host interaction, though the source does not say what that mode might be.
Comparing spacers across samples, particularly over time, showed how fast CRISPR arrays change. Microbial genomes remain largely stable, yet spacers were replaced rapidly, consistent with a defence system that must keep up with new threats. Even so, some spacers stayed unusually conserved across time and place, most likely because they defend against common and abundant viruses.
Why it matters
Knowing which viruses can enter which cells, and where they can complete an infection, is considered essential for future biotechnology that would use viruses to manipulate microbiomes. The work supports the DOE Biological and Environmental Research programme's aims for microbiome analysis and manipulation. It also contributes to the NIH Common Fund Human Virome Program's work on tools for viromics.
For researchers, the database provides a large reference for host prediction, a task that has relied on much smaller collections. The conserved spacers and the unexplained 10% of targeting are obvious candidates for follow-up, as are studies of the molecular features that let some viruses reach a broad range of microbial cells. The source gives no detail on validation of host predictions or on limits of the approach, which will matter as others begin to use the resource.
Source: Mapping the microbes' arms race with viruses at planetary scale (Phys.org, 21 November 2026).
Image: AI-generated illustration.
Working on this topic? Find a journal for your research or join the discussion on Scholars Post.


