The Origins of Mercury Lab
Updated: Mar 28
From 2014 to 2016, MRIGlobal was staffing a series of mobile BSL-3+ molecular diagnostic laboratories in Guinea and Sierra Leone during the West African Ebola Virus epidemic. MRIGlobal personnel ultimately processed over 19,000 samples, including nearly 100 positives, during their outbreak surveillance and diagnostics mission. While the teams worked tirelessly on rotating 12 hour shifts, it became clear that not all suspected EBOV samples were making their way to the laboratory. While the labs were ‘mobile’ in the sense that they could be put in the back of a transport plane and made operational anywhere in the world within 72 hours, once they were in country, they were a fixed-point lab. If it is rainy season in Sierra Leone, it is not likely that a suspected EBOV sample in Kambia is going to make it to a mobile BSL-3+ facility in Moyamba, nearly 200 km away. I saw the reasons for this myself in 2017, well after the outbreak had been subdued, when I co-led a molecular viral diagnostics training event for the Ministry of Health in Freetown, Sierra Leone.
Around the same time as the EBOV outbreak, several companies were emerging with the first hand-held molecular biology hardware that was 1) rugged enough, 2) simple enough, and 3) ‘affordable’ enough, such that a truly mobile, fully-outfitted molecular biosurveillance laboratory could be envisioned for routine deployment. In December 2015, I started as a post-doc in Jonathan Jacobs's group at MRIGlobal in Gaithersburg, MD. I don't think I had even been there a few days before Jonathan and I started seriously discussing these ideas and sketching out mock-ups on the whiteboard. We wanted to link quantitative real-time PCR with genomic sequencing, so that known and unknown targets could be interrogated in the field, outside of a brick-and-mortar laboratory. Biomeme Inc.’s two-3™ qPCR machine and the MinION from Oxford Nanopore Technologies Inc. provided a starting point for Mercury Lab version 0.1 (v0.1).
Mercury Lab v0.1
The goal was simple – a ‘lab in a backpack’. The two-3™ is the size of a coffee mug and the MinION is the size of a candy bar. Surely a backpack was all we would need for those devices and their ancillary consumables.
To test the concept, we teamed up with a pair of entomologists (Drs. Nathan Burkett-Cadena and Erik Blosser) from the Florida Medical Entomology Laboratory. Nathan and Erik were conducting studies on Culex cedecei mosquitoes, a vector for Everglades Virus (EVEV), and offered to have us tag along on one of their sampling trips. Our goal was to use our ‘lab in a backpack’ to conduct arbovirus surveillance for EVEV and determine if what we were envisioning would be truly effective in the field. We purchased an expeditionary photography backpack and packed it with everything we thought would be necessary.
Fieldwork has a tendency to illuminate that which you did not consider. Our Everglades trip was no exception. While we convinced ourselves that the hardware was viable (Russell et al. 2018), we found that a ‘lab-in-a-backpack’ was ill-equipped for all but the leanest of biosurveillance missions. To not only carry, but enable, the full complement of available hand-held genomics technology, we needed a more robust platform. Deep sequencing on the MinION exhausts a laptop battery fairly quickly. We needed more power. The library prep reagents and extracted nucleic acids needed better cold-chain if we were going to be in the field more than a few hours. We needed a stable workbench, preferably with some shelter from wind blown contaminants, rather than a rickety camping table (or a flat rock). We needed enough working area to set up multiple stages of our workflow and easy access to all of our pipettes and consumables, so we were not going back and forth into the backpack every other minute. We needed better computational capacity to run our preferred bioinformatics software. In short – we needed a laboratory.
Upon returning, Jonathan and I sat down with MRIGlobal engineers J.R. Aspinwall and Alex Boardman and began creating something in between a ‘lab-in-a-backpack’ and the mobile BSL-3+ laboratories we were staffing in Sierra Leone – something that took advantage of the small form-factor of modern genomics hardware, but did not compromise on logistics. We wanted a single, lockable, grab-and-go case that could be wheeled through an airport and checked as luggage, thrown in the back of a pickup truck, or even strapped to the back of a motorbike. We wanted something that could be set up on site in a few minutes or less, and be packed up just as quickly. We wanted a way to bring the lab to the sample.
After several prototypes (Mercury Lab v0.5), equally illuminating field excursions (Mercury Lab v0.5 in the field), and the input of specialist mechanical, electrical, and industrial design engineers from Bresslergroup in Philly, our team at MRIGlobal developed the first truly mobile, single-person portable molecular biosurveillance laboratory...
Mercury Lab v0.9
We developed two prototypes of the most comprehensive solution we could think of at the time, with a plan to spend time "kicking the tires" and identifying the inevitable gaps that we had missed. In the late summer of 2018, we took delivery of what I would now call our "v0.9" prototype. Within 2 weeks, it took its first plane ride as I demo'd it in Phoenix at the International Symposium on Human Identification. Later that Fall, I took it for demonstration, testing and feedback at NAMRU-6 (Naval Medical Research Unit No. 6) in Lima, Peru. A colleague took it to a coffee plantation in Colombia. I took it to a veterinary testing site in Nebraska. I demo'd it at USDA headquarters in Ames, IA and NIAID headquarters in Bethesda, MD and the National Renewable Energy Laboratory in Golden, CO, and Los Alamos National Laboratory in NM. In the Spring of 2019, I traveled to Addis Ababa, Ethiopia to include Mercury Lab v0.9 in an instruction module for a pharmacogenomics training event at the Ethiopian Biotechnology Institute. Throughout all of these interactions, we kept receiving very excited vibes and encouragement. At every stop, people nodded in agreement as we described our frustrations with currently-available field setups for molecular biosurveillance (i.e., stressful, ad-hoc, involving a lot of luggage and maybe a little bit of duct tape). It was clear to us that we had identified a widely-held frustration, and we were on our way to building something that may alleviate it. However, throughout our travels, we also identified those gaps we'd been looking for.
First, we needed to address power. Our v0.9 plan was to use the Midland PPG100 Portable Power Station -- a 950 Wh brick of lithium-ion battery power. While this product was easy to use "right out of the box" and would provide plenty of power for a substantial amount of field-forward sequencing, it did have drawbacks. When Samsung phones started catching fire on planes because of lithium-ion battery issues, no one wanted to take a 95o Wh brick of the stuff on their shipping flights. It was going to cost me more than three times the price of the battery pack itself just to ship it to Addis Ababa for our training event. There are no portable power supplies that can ship easily, or cheaply. We needed to find a way to allow our users to travel with the lab without a portable power supply, but have the lab work with a cheaply and widely available power source that they could expect to find locally at their destination, no matter what their destination was. While we considered solar, the size of the array needed to power the lab for the full range of possible utility was limiting in terms of "portability", and would not be particularly helpful for users that wanted to take their lab to often-cloudy or high-latitude destinations. Without question, the large-enough power source that ranks the highest in terms of being widely and cheaply available at almost any destination in the world is good ol' 12V lead-acid.....car batteries. So, we were going to need to figure out how to safely power Mercury Lab v1.0, and any sensitive electronics that users plug in, via the often volatile, 'dirty' 12V power source.
The next thing we needed to address was computational capacity. While Mercury v0.9 had a reasonably powered Intel NUC computer driving its analytics, it lacked the architecture necessary to perform some of the more rigorous computational work that is common to the use of devices like the MinION nanopore sequencer from ONT -- primarily, basecalling. The MinION device works by pulling strands of genomic material from a given biological sample through an array of tiny holes ('nanopores') on the surface of its flowcell. The flowcell has an electrical voltage running across it, and as nucleotides from genomic material transverse the nanopores, the voltage is disrupted. This voltage disruption occurs in predictable ways for series of specific nucleotides, and so we can 'back-calculate' what the sequence of nucleotides was from the genomic material by looking at the patterns of voltage disruption. This process, however, is very computationally demanding. It has been shown that the highly-parallelized architecture of GPUs (graphical processing units) is much more amenable to shouldering this workload than typical CPUs that you find in most laptops and consumer computers. While ONT makes an accessory device (the MinIT) that leverages GPUs to conduct basecalling, we did not have an extra USB port to plug it into with the way the NUC was configured in the monitor housing on Mercury Lab v0.9. Additionally, we wanted to give our users even more juice than the MinIT. Since fieldwork is often done in austere conditions that may not necessarily be comfortable, we imagined "lengthy-waits-for-basecalled-data-while-swatting-mosquitoes" being a future frustration that we would ultimately end up fixing, so why not just build in the extra capacity now. Plus, with substantially more CUDA cores, users would have the flexibility to train their own, more-accurate basecallers for their specific biological systems (if they ever deemed it necessary) directly in the field. All of this added capacity was going to ensure we would need to re-design the computer housing, which was serendipitous since I really didn't like the wobbly way the top-heavy monitor/NUC assembly sat on the corner of the laboratory, nor the way it stored during transit (we had two monitor/NUC assembly housings crack during travel).
Next, we focused on improving the laboratory workbench itself. The v0.9 workbench was made from a PVC/acrylic blend called Kydex, and swung into place from a hinge underneath the soft-sided storage container/removable-backpack that served as our laboratory shelving. The far side of the table was supported by two telescoping legs, hinged underneath the far side corners. Selected for its thermoformability and resistance to chemicals (thus, amenable to routine sterilization), the Kydex bench was heavy and the set-up process felt clumsy and unwieldy. It was up to the user to keep adjusting the telescoping legs back and forth to achieve an even bench, and the leg hinges did not withstand a heavy beating very well. We needed something lighter, but still durable -- something that could be stored more securely, but set-up just as fast -- something that added rigidity and stability, but did not compromise on utility.
All of these changes were going to require *some* compromise. We were trying to build a lot of capacity into a small space and there was not much room left to play with. But, as anyone actively engaged in the genomics research community for the last few years will confirm, the most steady constant is the rapid pace of evolving technology. While this can often make designing a product to service a need in the community feel a little like an archer trying to hit a bullseye mounted on a bullet train, a particularly significant change came about at just the right time for our design timeline. Mercury v0.9 held two separate cold storage containers -- CREDO cubes that can hold approximately 12L of items, either frozen at -20ºC or refrigerated at 4ºC for ~96 hours. However, ONT developed a "field kit" of sequencing library prep reagents that was stable at room temperature. This effectively eliminated the need for -20ºC cold-chain storage and allowed us to recoup a substantial amount of weight and space to play with in design.
A laundry list of other items remained -- more handles for carrying, larger wheels to more effectively traverse rough terrain, a better windscreen, an illumination source for work in low-light conditions, etc, etc. We were lucky to have Alex Terray, with his substantial engineering and microfluidics background, join the project. His council on design issues has been critical as we checked off this laundry list.
Over the past 6+ months, after the last of the tire kicking, we have been working towards our 'go-to-market' version of Mercury Lab...the version that we could feel confident in and stand behind, knowing that it was a good answer to many of the frustrations experienced by anyone attempting genomics-based data collection in austere, far-afield locations.
After a very long, meandering road, I am very excited for the future of this platform, and grateful for the friends and experiences I have gained along its development path.
Onward...to Mercury Lab v1.0.