Which emerging technology detection platforms are a good fit for my needs?
Updated: Jul 8, 2018
A friend asked me this question after a recent discussion at a biodefense conference. This has been a bet-your-career kind of decision dilemma for me for the past ~20 years. Welcome to my nightmare!
Choosing the “best platform” is a difficult multi-dimensional product-market fit optimization problem involving many variables, each with potential impedance mismatches between user requirements/constraints and platform features. This includes but is not limited to factors such as:
Assay depth needed to cover the desired threats, versus assay depth available on the platforms under consideration.
Degree of nucleotide strain variation among the species needing to be covered (roughly, higher strain variation can mean more unique assays are required to cover all strains.)
Time-to-answer desired, versus that of the available platforms (workflow).
Reagent cost/sample target desired by the end-user.
Degree of sample multiplexing needed and that of the available platforms (throughput).
Sensitivity and specificity required by the end-user and that of the available platforms.
Sample type(s) required. (blood, saliva, urine, feces, CSF, tissue, swab, soil, aerosol, food, product, water, etc.) Some platforms handle only limited sample types. Some pathogens are only detectable in certain clinical sample types.
Degree of automation of sample prep and nucleic acid extraction required for desired sample type(s).
Whether both RNA and DNA need to be processed from the same sample (may require making cDNA or else splitting the sample into independent RNA and DNA analyses.)
Whether or not sporulating agents need to be detected (need sonication to crack spores to extract DNA, instead of simpler chemical cell lysis.)
Skill required of operators to run the assay (affects training and labor costs.)
Bioinformatics infrastructure required to analyze data in required timeframe (local or cloud processing; whether local bioinformatics and/or IT staff are required, etc.)
Skill required of end-users to interpret results properly.
Footprint of the device and other operating characteristics (e.g., power, cooling or other physical requirements for operation.)
Purchase cost of instrument and cost of annual maintenance contract.
Ruggedness required by end-users and that of available platforms.
Degree of maturity of platforms; degree of stability of their manufacturers.
Acceptance of the technology by end-user community and any pertinent regulatory agencies.
Can user supply their own assays or probes? (for technologies where Open Access is applicable.)
Getting reliable answers from vendors for platforms, especially new ones, can be tricky. Comments found on the Web, whether favorable or unfavorable, must be taken with a large grain of salt. Understanding the pathogens that need to be covered is also tricky: do you need to reliably detect EVERY known strain of each target species, or can you limit to those known to be currently circulating worldwide (or those in a specified region)? This is especially important for RNA viruses: you probably don’t need to worry about a strain last seen in the 1950s (unless you are concerned about potential bioterrorism that might have access to historical strain collections.)
Most of us can’t optimize in our heads the kind of high-dimensional space represented by the many variables listed above. Instead, we typically determine the 3-5 key requirements/constraints and try to choose from the available technologies to best fit them. Example of key requirements/constraints might include:
“Total workflow time from sample-in to results must be no longer than 8 hours.”
“Total reagent cost/sample must be < $N.”
“System must process blood, saliva, or urine.”
“Sample volume may be as small as N ul.”
“Limit of Detection should be between 100-1000 genome copies/ul.”
“All our sites already have MiSeq instruments, so we have to use that platform.”
Note that the last example listed is not a true user requirement; instead, it is an attempt at a non-technical constraint due to perceived cultural or political mandates, regardless of whether that platform is indeed optimal or not for the task. These kinds of political constraints are unfortunately all too common in practice. In my opinion, the proper response to them is along the lines of “My analysis will explain which technology best matches your requirements; you can then decide whether you will accept that result or not.” Beware that this response is not always received gracefully!
Sometimes you will find that the user requirements turn out to be impossible to fully satisfy with current technologies. Everyone seems to want some version of the Star Trek TriCorder, so you frequently must negotiate their initial requirements into something that can actually be achieved with the technologies of today. Asking for a DNA sequencer that is infinitely fast and free and sequences everything in your sample and requires no bioinformatics sounds like the annual request from the Defense Department for the TriCorder but is not something that will be achieved any time soon. (Regardless what some Oxford Nanopore fanboy may tell you.)
Picking detection technology “winners” can at times seem more art than science. Your best bet is to try to examine as many variables as possible, understand which user requirements are truly critical, and evaluate the range of technologies with an open mind about their true technical abilities and limitations. Not every problem is best solved with sequencing, or PCR, or mass spec, or microarrays; but if you handle enough use cases, you will eventually find that almost every detection technology class is a “winner” for at least some combination of user requirements. Good luck!