You don’t need robots and computers, just a system and apparatus to organize the steps!
This model is based on the results alluded too but not completely explained by an Israeli company in the recent NY Times article by David Halbfinger on creative pool testing options. https://www.nytimes.com/2020/08/21/health/fast-coronavirus-testing-israel.html
Trying to come up with one of these is a great activity for teachers, students or anyone who wants to understand testing strategies better. And it’s not that hard. We want a pooled test that can give more test results for less lab chemicals and time. At the extreme, lab engineers in Israel and other places are claiming to return 8 results for 1 test, or even 9 results for 1 test.
We set out to figure it out, build an apparatus and test it out using vinegar, water and PH test strips. There are a lot of options with different features. We did a frame for simple pooling (see photo), but then we really went for it (scroll down) with a combination of two matrices and overlapping pools. It totally worked, no robotics or computer controlling the pipette work which took 27 minutes. We’ll try to shoot a video and I bet that time comes down.
SIMPLE POOLING For one infection in the group, 10 tests: 25 results Here’s a Video (less than 2 minutes) to show how this works.
COMPLEX MATRIX AND OVERLAPPING POOL, 8 tests,72 results
This strategy works if there is a low prevalence. We are expecting 0 or 1 infections out of 72.
It’s a 2 part design. A 6 x 6 matrix to put the samples into sets of 12 pools, 6 vertical, and 6 horizontal. There will be a positive on the associated vertical and horizontal pools so do a second matrix and we mix and match on the next step (useful insight). We take the vertical pools together and using the overlap guide, place them into 4 groups I, II, III, IV. We did the same for the horizontal. We now have just 8 tubes to test, and the results show us which row and column we should reference to find the vinegar sample back in the matrices. Actually doing the experiment was unexpectedly fun (see bottom).
It worked! We did a hidden vial of vinegar test. Taking 3 drops from each vial and putting them into into the matrix pools (vertical and horizontal) so each sample goes into two vials on step 1. Then we used the overlapping pools guide to help organize 12 samples (both verticals) into 4 vials. Which vials are positive for vinegar, 1st, 2nd, 3rd or 4th or 1st and 2nd, 2nd and 3rd, 1st 2nd and 3rd, etc tells us which single pool has the vinegar, and that pool tells us a column in one of the two matrices. When we do the horizontal pools together, we expect the positive to come from the same matrix as the vertical (or there’s a problem). The one sample that is in both the positive vertical pool and the positive horizontal pool is the one we are looking for and its at the intersection of the two pools we identified.
Having more than one positive sample could mess this up, but if you could use this method, for less than 27 minutes of pipette time you can leverage your lab equipment 1:9. If there are multiple positives, you would either test all 24 pools in the matrices or test all the samples. Positives in one side may mask positives in the other so it may be smart to switch the position of the horizontals or think strategically about the masking issue.
Other constructs avoid this problem, but may cost more tests. Here’s one I’m tinkering with a 9:36 or 1:4 ratio of tests to results. The top half and the bottom half make sense, the middle is trials at matrix after overlapping, but I don’t think that works as well as the other way. Bottom right I’m trying to figure out how many samples could be in question if different numbers of the pools test positive.
Working on posting Videos just as a record. First try of the experiment worked perfectly, the video recording not so much, but here are some if you care to look.
Next up, tackle the Rwanda Hypercube. This is 3 x 3 x 3 cubes, and then multiples of them. The method described in this article also includes a surveillance pool that creates a high probability of 1 positive with a low probability of 2 for a given number of samples. With that preparation, the right matrix, cube or hypercube can be selected to give the most efficient results. https://arxiv.org/pdf/2004.14934.pdf