Frequently Asked Questions
Are pooling and multiplexing cutting edge, new methods?
No, not at all. Pooling and multiplexing are widely used in science and engineering.
Dorfman in 1943
published one of foundational works in group testing, and since then pooling and multiplex assays have been a standard method for quality control and experimental design. Pooling assays in public health were first widely used for screening pathogens in blood donors. More generally, multiplexing is a core part of information theory that is the basis of digital communications and computing.
If you are interested in looking at a much larger set of pooling designs, you might look at our sister project Disjunct Dictionary
Is pooling used for pathogen screening currently? Where?
Yes, pooling of biological samples is widely used, in particular for pathogen detection. As an example, the American Red Cross routinely uses pools of 16 samples to detect hepatitis B, Zika virus, and the Babesiosis (Babesia micoti)
. In all of theses cases, the disease is rare, but if a positive pool is detected then the individual samples are retested and sent on for further analysis.
Are other people looking at using group testing for COVID? Where?
Yes, Origami Assays is far from alone. For a long list of examples see our page on examples of pooling.
If you know of others, let Peter us know (firstname.lastname@example.org).
Can COVID19 samples be successfully pooled?
Yes, this has been demonstrated by the following groups:
In general, an RT-PCR assay requires a small sample volume (~5 microliters), while the patient sample is much larger, on the order of 2 milliliters--a 400 fold excess.
Pooling is most easily done before RNA purification. RNA purification requires approximately 100 microliters of the patient sample. If there were 20 samples per assay, then each RNA pool would include 5 microliters of patient samples. A benefit of pooling before RNA purification is that the pooling then reduces the number of RNA prep kits required and keeps all of the liquid handling steps a fairly easy size.
Is multiplexing used for COVID19 screening? If not, why not?
It might be in some places, but often not. As of June, 2020, the largest use I've found of nonadaptive group testing for COVID-19 is from Rwanda.
Based on this interview, of the 168,000 tests performed there to date, 50% have been done with pooled samples.
I suspect the reason multiplexing is less often used is because it is a different way of thinking and somewhat complicated to run. Most of us are familiar with the idea of running one assay to get one result--this makes sense and fits our intuition. However, running 10 multiplexed assays to get 45 results (the S2 design
) seems like magic.
When multiplexing is used, the mixtures are made by robotic liquid handlers to ensure that the right mixtures are made. Because robots excel at tedious precision work, they are well suited to very large multiplexing assay design.
However, we believe that smaller multiplex designs are still useful and do-able by human hands if given appropriate structure.
Another concern about pooling in particular
is that it can amplify false negative calls, particularly when no replicate testing is done. In Origami Assays, all of our multiplex designs contain a minimum of two fold replication that overcomes this potential for false negative results.
Multiplexing works best with "needle in a haystack" type problems, meaning assay situations where the prevalence rate (expected positive rate) is low. In these cases, the Origami Assay multiplex designs give robust results with many fewer assays even in the face of moderate assay error rates. However, if the prevalence rate is beyond what the multiplex design can handle, the assay will fail gracefully in that it will call additional false positives. Failing by calling false positives is preferred for pathogen detection because the number of positive calls are far fewer than the negative calls, making it possible to weed the false positives out by a second validation round of tests on the putative positive calls.
What is the difference between pooling and multiplexing?
The two terms are related:
- Pooling: Mixing multiple patient samples together in a single assay well.
- Multiplexing: Constructing multiple different pools such that the combined assay results yield a maximum amount of useful information.
Simple pooling is when instead of running one sample per assay, a group of samples are mixed together and run on a single test. In this simple pooling scheme, if a pool is negative the sample is assumed negative. Simple pooling strategies are the ones widely used by the American Red Cross
for blood screening as noted above.
Multiplexing creates a set of pools where each sample may be tested multiple times in different mixtures. This additional testing can provide robustness to assay errors and/or greater assay precision. Common examples of multiplex designs are the Reed-Solomon coding schemes
used in CDs, DVDs, and QR codes. These multiplex designs reformat data in different ways to better handle errors and/or provide data compression. The designs in Origami Assays uses a similar approach to generate compression and error tolerance schemes for biological assays.
Can origami assays be run using something different than a 96 well plate?
Absolutely. The multiplexing designs used in origami asasys are general. If you have an assay that isn’t in a 96 well plate, you can simply label each assay tube or well, wherever it may be, with the letter and number combination used in the origami assay software.
What if I don’t have enough samples to fill a whole assay?
If you are close, but don’t have quite enough samples to fill a whole assay, that is fine, just skip the sample templates that you don’t use. The result will be that the assay has less compression, but it will also tend to produce fewer false positives.
After you decode the asasy, if you see that some of the skipped samples are called as positives, then you can safely eliminate these as false positive results.
How reliable are multiplexed assays?
Multiplexed assays are both more robust and more fragile than monoplexed assays. Multiplexed assays are more robust in that they test each sample multiple times, so that if one well were to spontaneously fail, the positive result could still be detectable. However, to detect this positive result requires a low true positive result and a modified decoder that calls near perfect hits as positive results.
Multiplexed assays are more fragile in that they use a smaller volume of sample and are subject to more human error. Ideally the assay should be very sensitive to even the smallest amount of virus for example, but all assays have a lower detection limit. Furthermore, accurately pipetting a multiplexed assay is challenging and can potentially introduce error.
However, monoplexed assays also are fragile. In a monoplexed assay, each sample is run only once, so if the well fails then the result is incorrectly called--there is only one chance to get it right. Similarly, human error is also a problem for monoplexed assays because they involve running many times the number of assays for the same amount of data.
Are multiplexed assays approved by the US FDA?
Apparently the FDA doesn't weigh in on this kind of design. We asked, and the FDA's response was: "the agency does not establish nor provide input on population testing protocols such as your proposed sample pooling strategies."
That said, there have been at least two Emergency Use Authorization applications for different pooling schemes:
Note that updated HHS regulatory changes
(as of August, 2020), have largely removed federal oversight for COVID-19 testing.
What does the internal consistency check for the decoded result mean?
When decoding an assay result, Origami Assays doesn't assume that everything went correctly. There could be cases where an assay failed or the design was pipetted incorrectly. If this were a monoplex assay these errors could not be identified, but for a multiplex assay these errors are detectable.
In the Origami Assay decoder, we first identify what the most likely pattern of positive results was based on the assay result. With this call in hand, we then go back and simulate what assay result we would expect given those positive results. If the simulated results don't fully agree with the observed assay result, we note that inconsistency in the decoder.
If a decoding is found to be inconsistent, we recommend retesting all of the called positive and possibly positive results. We also recommend checking your assay to make sure it is performing well.
What does a “called positive” result mean? And what does the star ("*") mean?
The ultimate goal is to define if each sample is positive or negative based on the multiplexed assay result. However, there is a bit of nuance to this answer.
Broadly, a positive result is a result that is consistent with the assay result, while a negative result is inconsistent with the assay result. That is the easy part.
However, because the assay is multiplexed, we can differentiate between some positive results that are necessary (marked by a star “*”) and others that are not necessary but are still consistent with the assay results. Necessary positive samples are samples that must be positive to produce at least one of the assay results, meaning there is no other way to “explain them away”.
As an example, consider a case where four samples (1, 2, 3, and 4) are distributed in paired sets into 5 assay wells (A, B, C, D, E) as is shown in the table below:
Imagine that we observe that wells A, B, C, and D are positive and E is negative. When decoding, we find that samples 1, 2, and 3 are all counted twice (both pairs are in positive wells), so all three samples would be marked as positive.
However, samples 1 and 3 would be marked as 1* and 3* because they are necessary. To show this, imagine if sample 1 were negative, then well A would be negative, which would be inconsistent with the observed assay result. Similarly, if sample 3 were negative, then assay D would be negative, which is also inconsistent with the observed assay result.
Sample 2 being positive is consistent but not necessary for the assay result. If sample 2 happened to be negative, then it would not impact the assay result assuming 1 and 3 were positive. In effect, the results of sample 1 and 3 hide the status of sample 2 so we can’t say if it is or is not positive based on the assay.
Note that in larger multiplexed arrays with more positive assay results, we often find that none of the samples are listed as necessary. This just means that there are many ways to potentially explain the assay results, and none of those explanations all require any one sample to be positive.
How do you make money off of this?
I don’t make anything off of origami assays. Multiplexing is a technique that has the potential to expand our ability to test for COVID19. More testing is extremely valuable to all of us, so I'm putting this tool out for everyone to freely use.