Coronavirus detection via PCR.... is this idea feasible?
March 6, 2020 8:14 PM   Subscribe

Given the difficulty we seem to be having testing samples right now (see, e.g. this article), I'm wondering if this little trick might increase our capacity by a pretty large factor. PCR experts, is this possible?

Couldn't one commingle a bunch of patient samples and test the commingled material? If it comes out negative, everyone's cleared; if it comes out positive, you test individually to see which one(s) are truly positive and which are negative.

So long as there are more negative samples than positive ones, you'll wind up increasing your capacity by a significant fraction. (Example with numbers below.)

My question: given the way RT-PCR works, is this feasible? Could one run the test on a bunch of commingled samples like this and still get a reasonable reliability for a positive/negative on the group sample?

(Example with numbers: assume 1 in 16 samples on average is positive. Instead of testing each sample one by one, commingle 8 samples and test that. About half the time -- actually, closer to 60% -- all eight would be negative, and therefore the commingled sample would come out negative: one test for 8 patients in all, and you're done. If the commingled sample comes out positive, then you test each of the 8 patient samples individually: nine tests for 8 patients.

The gains by only needing one test half the time outweigh the loss by needing an extra test half the time -- if my math is right, it's like needing only 5 tests for every 8 samples, on average. My question is whether this is feasable in the real world. Of course, assume that there's enough genetic material in a sample to run at least 2 tests for each patient.)
posted by cgs06 to Science & Nature (7 answers total) 3 users marked this as a favorite
I am dumb. But This Week in Virology takes audience question.
posted by j_curiouser at 9:06 PM on March 6, 2020 [2 favorites]

Best answer: You can definitely do this kind of efficient pooling, at the expense of turnaround time, since you have to retest the samples that make up the positive pool. I think deploying automation is a more effective way to get larger screening capacity - setting up RT-PCR reactions in 96 or 384 well plates is not that hard.
posted by pombe at 9:14 PM on March 6, 2020 [1 favorite]

Sample tracking is also more complex in this case as well.
posted by pombe at 9:14 PM on March 6, 2020 [1 favorite]

Best answer: I would be concerned about false negatives for samples with viral RNA near the assay's lower limit of detection, as they'd be diluted by a factor of 8 if pooled with seven other negative samples as in the example. The LOD in this published nCOV assay is a little under 4 gene copies per reaction and the 1/8 pooling would be dropping it down to <1 (or a high chance of getting zero).
posted by fountainofdoubt at 10:16 PM on March 6, 2020 [11 favorites]

The CDC is using real-time RT-PCR for diagnostic purposes currently.

See some discussion here and the documents on the CDC website here, particularly the one titled "CDC 2019-nCoV Real-Time RT-PCR Diagnostic Panel Instructions for Use." There's a section named "Limitations," which is relevant to what fountainofdoubt mentioned:

Negative results do not preclude 2019-nCoV infection and should not be used as the sole basis for treatment or other patient management decisions. Optimum specimen types and timing for peak viral levels during infections caused by 2019-nCoV have not been determined. Collection of multiple specimens (types and time points) from the same patient may be necessary to detect the virus.
posted by typify at 11:09 AM on March 7, 2020

Best answer: Via Joe in Australia: To ease global virus test bottleneck, Israeli scientists suggest pooling samples
posted by mbrubeck at 7:33 AM on March 20, 2020

« Older Consultant-filter: when "collaborate with X" means...   |   Need to go bananas... Newer »
This thread is closed to new comments.