An Update on Our Air Sampling Research
Cross-post from the NAO blog.
We’re pleased to announce that our review on indoor air sampling for detecting viral nucleic acids has been published in the Journal of Aerosol Science. This peer-reviewed publication builds upon the preprint we previously announced in May 2024. It covers:
The sources and composition of viral bioaerosols in indoor environments, including pathogen concentration data from qPCR studies as well as relative abundance data from metagenomic sequencing studies.
Methods for sampling viruses, including using dedicated collection devices with air movers, settled dust, and HVAC systems.
Strategies for effective implementation of air sampling for biosurveillance, with a focus on monitoring the air travel network and healthcare settings.
We believe the article represents the most comprehensive resource on indoor air sampling for viruses to date, with relevancy to a variety of audiences and applications, including public health, biodefense and biosecurity, aerobiology, and industrial hygiene.
At the NAO, our primary mission is developing early warning systems capable of detecting novel pathogens before they spread widely. While our review adopts a somewhat wider scope in order to be useful and interpretable to other applications, it’s worth revisiting our thoughts on air sampling in the context of the NAO’s mission.
When we set out to investigate air sampling, it was part of a broader effort to evaluate different sampling strategies. We wanted to ensure our focus on wastewater surveillance was evidence-based rather than the result of early assumptions that hadn’t been thoroughly tested. Since then, in addition to scaling up our wastewater surveillance program, we’ve performed in-depth investigations into indoor air sampling, blood-based biosurveillance, and nasal swab sampling (the latter of which has progressed to an active pilot program).
So what’s the state of our thinking on air sampling in relation to other approaches now?
We believe that indoor air sampling could offer valuable complementary detection capabilities to wastewater but isn’t sufficiently promising for us to divert resources to sample collection ourselves at this time. However, we remain excited about the potential of air sampling approaches more broadly. Our assessment not to prioritize air sample collection is based on several key considerations:
Viral bioaerosol sampling is challenging. Many air-sampling approaches struggle with capturing the small particle sizes associated with airborne viruses and collecting enough viral material for sequencing. Promising alternatives like HVAC filter sampling or settled dust collection lack data on their effectiveness at collecting viral material for analysis. While there are certainly promising examples of viral bioaerosol sampling (for example, work from David O’Connor’s lab), our team currently lacks operational experience with air sampling and we anticipate that building a strong competency would be quite resource-intensive. Unlike the relatively straightforward process of getting wastewater samples shipped from treatment plants, we currently don’t have an established pathway for accessing or collecting quality air samples at scale.
Air sampling tends to have smaller catchment sizes than wastewater. The number of people contributing to an indoor air sample is typically much smaller than those contributing to a wastewater sample from a treatment plant. While specialized populations (like international travelers) can still make smaller samples valuable, we generally expect larger catchment sizes to provide superior detection capability. Scaling air sampling probably means deploying more samplers, which quickly multiplies both cost and logistical challenges.
Air sampling may not be as effective at capturing respiratory viruses as hoped. One of our main reasons for exploring air sampling was its theoretical advantage in detecting airborne respiratory pathogens—a class of pathogens we think are particularly likely to cause future pandemics. But there is some evidence to suggest that air sampling may not be as promising for this purpose as we hoped. In metagenomic studies of air samples, human-infecting viruses typically make up only a tiny fraction of reads (similar to wastewater), with most coming from skin-associated pathogens rather than respiratory ones. Results from Rosario et al. 2018 and Prussin et al. 2019 as well as our own independent analysis1 of the sequencing data showed minimal content from common human respiratory pathogens, typically appearing in fewer than 15% of samples with relative abundances2 below 1e-6. Minor et al. 2023 did better with SISPA amplification and deep long-read sequencing, but we’re still uncertain about how effective air sampling will be for respiratory pathogens when paired with metagenomic sequencing for biosurveillance.
Despite these challenges, we’re still interested in air sampling’s potential and are open to supporting promising efforts in this space. We’d be especially excited about collaborations where we could run our pathogen detection algorithms on data collected by partners with air sampling expertise. In particularly promising cases, we’d even consider covering sequencing and sample processing costs to make this happen.
As always, please reach out if anything here is interesting to you!
Footnotes
We presented our air sampling analysis as a poster at CBD S&T 2024, which includes species-level taxonomic assignments created with our mgs-workflow pipeline (v2.2.1). This pipeline performs taxonomic assignments using Kraken2 and Bowtie2, as described in our third blood biosurveillance blog post. We think this approach is not optimal for making high-fidelity species-level assignments (Fig 4. in the poster), but we believe the findings are robust enough to support our general conclusions about air sampling presented here.↩︎
It’s worth noting that both Rosario et al. 2018 and Prussin et al. 2019 used relatively shallow sequencing depth, which impacts the detectable range of relative abundances. With deeper sequencing, it’s possible to detect lower-abundance species that would otherwise be missed, extending the lower range of observable non-zero relative abundance values.↩︎