At-Home COVID Testing for Pirates

In the age of COVID, we are all worried all the time. Am I sick? Did he just breathe on me? Will I kill my parents? I’m not even worried for myself, I say, but am I going to be the asshole that infects everyone? Every decision feels so heavy1 all of the time. It’s exhausting.

A friend told me one day: a group of friends planned a get-together, and everyone got tested in the morning at a nearby center with a turnaround of a couple hours.

Brilliant. Testing. How obvious a salve! If I could know with reasonable confidence before and after an event that I was safe, I wouldn’t have to worry so much about killing my friends and parents, and I could do more things and see more humans.

But, of course, in my area, I couldn’t find a testing center that would even test me without symptoms, much less give a sub-day turnaround. And, given the timeline of infection, I would really want frequent access to testing to make sure I wasn’t missing the onset of an infection. Alas.

I admit I raged against the machine a little— how can we not have better testing infrastructure at this point2? Why isn’t every PCR in the country running around the clock? Heck, I ran tests like that all the time in the lab— how hard could it be?

“How hard could it be?” — famous last words in tech. But I decided to find out, and it turns out that the answer is… not that hard. It’s actually surprisingly easy to set up, and I now have an at-home COVID testing rig that I have been using to test myself and my family. There are many caveats to that statement, and I will get to those, but, first— let me tell you a little bit more about how I set everything up.

What is a PCR, anyway?

There are a number of new technologies being developed to make testing for COVID easier, faster, and cheaper3, but the gold-standard is still a PCR-based test that looks for fragments of RNA from the virus. PCR stands for “polymerase chain reaction,” and it refers to the well-established process by which scientists can make copies of a given piece of DNA: a sample containing the target DNA, special DNA fragments called “primers,” and special enzymes are run through a temperature cycling program that replicates the target DNA sequence, doubling the number of target fragments with each pass through the program.

Through a thermal cycling program, DNA is replicated. The magic of science. 

PCR is run on a PCR machine. The simplest versions of these are thermal cyclers: there are little holes to put lab-sized plastic tubes in, and the machine runs programmed temperature patterns. These machines are used extensively in labs and clinics to create libraries of DNA fragments being studied for any number of reasons.

A variation on the PCR theme is what’s called a “real-time PCR machine”4: a thermal cycler with the ability to measure fluorescence. Instead of just replicating the DNA fragment, a real-time PCR replicates the fragment via its thermal program while also measuring the amount of fluorescent signal created by the process of replicating each fragment in the presence of special fluorescent molecules5. A real-time PCR machine thus enables scientists to run a process called quantitative PCR, or qPCR, which answers the critical question, “How much of a given DNA fragment exists in this sample?”

The CDC protocol for COVID testing relies on qPCR: collect a sample from the patient, extract the DNA and RNA from the sample, add special DNA fragments called “primers” that match the genetic sequence of SARS-CoV-2, and run qPCR to measure the amount of DNA in the sample that matched the SARS-CoV-2 primers.

So, clearly, the first thing I needed to run an at-home COVID test was a real-time PCR. There are a number of major manufacturers, including Roche and Bio-Rad, but new top-of-the-line real-time PCR machines cost upwards of $20,000, which is more than I am willing to commit. There are also manufacturers in China that sell on Alibaba, but I couldn’t tell which machines would actually be real-time PCRs with software sufficiently in English for me to operate. Plus, after contacting several companies, it seemed they were reticent to ship to the US given regulations anyhow.

Which left the resale market. I was skeptical I could find anything given what I assumed would be a worldwide shortfall of PCRs, but I was happy to discover that the market for used real-time PCRs is robust: in addition to dedicated used lab supply retailers, there are a wide range of systems available on eBay. I watched the inventory cycle for a while, and soon enough found a good deal on the machine I wanted, the Applied Biosystems StepOne6. That was the machine I had used in the lab when getting my PhD, so I knew that it had the capabilities I needed and that the software would be ye olde and Windows-only, but usable.

Worth the $1,000 for shipping. 

A lab to call my own

And so, for $5,000, I became the happy owner of a real real-time PCR. COVID-testing aside, real-time PCRs are workhorses of genomics labs, useful for any number of experiments, so I figured if nothing else, I was one step closer to having my own private lab space. But, reading through the CDC protocol, I realized how much standard equipment is assumed in a lab or clinic: freezer, centrifuge, eppies, tube racks, pipettes, ethanol, distilled water, and much more, even before you get to actual reagents (that is, the consumable chemicals).

Luckily, eBay and Amazon are here to help. I was quite impressed by how easy it was to quickly stock my lab with everything from pipette tips to saline. The centrifuge and pipettes I bought used; I first tried cheap new pipettes from Alibaba, but they were clearly inaccurate and inconsistent. That left the reagents and consumables. Most, including the expensive enzymes required for qPCR, came from ThermoFisher, and the rest came from any number of lab supply companies that turn up in Google searches.

All in all, I got everything I needed for less than $8,000 (the final supply list is here). Finding a place to put it all was tricky, especially while we’re all in our apartment all the time, but I had one dresser’s worth of space to spare. And so, in our guest bedroom above our linens and next to our assortment of exercise equipment, I now have my very own COVID-testing facility.

Don’t mind the hand towels I’m using to line the dresser.

Running an at-home COVID test

It takes me about two hours of hands-on time to collect and prep samples, plus an hour to run the PCR and analyze results. I’m intentionally not going to go through the details of the testing protocol itself; I used the published CDC protocol, and if you are not familiar enough with RNA extraction and qPCR to follow that protocol, then I can’t be held responsible for what you do. But, at a high level:

  1. Nasal swabbing seems gross, so I collect a throat swab. This is the most dangerous part of the process for me as the “clinician”; reaching into the mouth of a potentially infected person is foolhardy. So, I only run the test for people who I am willing to die for7, and I have people self-swab for good measure.
  2. I use an Invitrogen viral extraction kit to lyse cells and collect RNA and DNA.
  3. Starting at the cell lysis step, I introduce a SARS-CoV-2 positive control (a fragment of the SARS-CoV-2 genome as a DNA plasmid) alongside the real human samples.
  4. I run reverse transcription (the process of turning RNA, either from the patient or the virus, into DNA) and qPCR in one go with a TaqMan 1-step master mix and each of three primers: two for SARS-CoV-2 genome targets and one negative control (a common human gene, RnaseP). I typically run two replicates per primer per sample to guard against my own imprecision.
  5. In theory, qPCR would allow me to run comparative analyses, to determine how much SARS-CoV-2 is detectable in samples relative to the internal controls. However, all I really care about is the binary result, as ANY positive signal for the SARS-CoV-2 would warrant an official clinical test.
We have a lax shoe policy in our lab.
The software runs in Parallels on my Mac laptop, so that I can watch the happy rainbows of successful qPCR form as the program runs.
The first two sets of bars represent signal from SARS-CoV-2 genes, while the third set is a common human gene used as a control. The y-axis represents the number of cycles of the thermal program required to detect signal for the target sequence, with bars at zeroish for no signal detected. The hardest part was getting Google Sheets to draw error bars, which proved impossible. Luckily, I only need the binary answer.

How reliable is testing?

I have now run a number of tests for myself and my family. Notably, I have not tested a COVID-positive person so far as I know, so I am not sure yet if any of this works at all. But the internal controls give me some confidence; I figure if I can successfully scrape human RNA off the inside of your throat, but not any viral RNA, chances are you’re not shedding a ton of viral particles in any case.

That said— I’m doing this in my house. Barefoot. Even if it were 100% accurate, there would still be a fourish day incubation window to contend with, during which someone could test negative but still be a threat. And I am of course not 100% accurate; even the official clinical tests have an error rate, and can return false positives (that is, return a positive result when the patient is uninfected) or false negatives (that is, return a negative result when the patient is infected).

My lower-bound estimate is that my test is 90% accurate, so I use the 10% chance I am wrong to scale the base risk percentage. For example:

  • Marla is very conservative about her COVID risk, and for the two weeks prior to testing, she goes to the market twice, with a mask. She has not experienced any symptoms of COVID in relevant memory. We can ball-park Marla’s risk as 0.1% chance she has COVID. If we additionally test her and she comes back negative, there is 10% chance the test is wrong, so 0.1 * 0.001 = 0.0001, or an adjusted risk of 0.01% chance that she has COVID after testing.
  • Maurice is a childless millennial, and he has been hanging out in opium dens as per his usual for the last two weeks. In the three days prior to testing, he learns that his bosom buddy, who he was driving with for hours, just tested positive for COVID. We can ball-park Maurice’s risk at 50% chance he has COVID, because, who knows. If we additionally test him and he comes back negative, there is 10% chance the test is wrong, so 0.1 * 0.5 = 0.05, or an adjusted risk of 5% chance that he has COVID after testing.

All of this is hand-wavy, but it gives me a framework for thinking about relative risk. If I were in Maurice’s shoes, 5% chance means that 1 in 20 times, I would have COVID, regardless of the test result. I wouldn’t see my parents with that risk. But if I am conservative in my habits without being maniacal, the test reduces my anxiety on any given day by an order of magnitude, which is often enough to make seeing my humans possible.

How has this changed my routines?

If I’m being honest, the real reason I did this was because it’s fun, I wanted a COVID hobby, and now I’m one step closer to being a famous scientist doing genomics research on a yacht. That said, it really has reduced my worry by an order of magnitude, because now I can go to restaurants (outdoors), visit other cities (by car), and even walk through diffuse crowds (mask on) knowing that as long as I give myself a fourish day buffer and a test, I am unlikely to carry anything to my parents when I see them.

And in that way, I went from seeing mostly my boring white walls all day, every day, to having weekend plans to look forward to. Having things to look forward to has been a fabulous release from the heaviness of this time; to be able to go out, see humans, and enjoy fall, knowing that I can easily alleviate my worry as necessary— what a blessing!

It should be easy. Afterall, all it took for me was … five years of a PhD and $8,000 in capital investment, plus two hours of hands-on time with every test.

Oh. Wait. The pursuits of the wealthy and privileged. Right. If only we had some sort of nationwide system that helped distribute access to The Future across the population. Sigh. In the meantime, I think my pirate lab needs some Influenza primers.


  1. In the Milan Kundera sense; every decision feels so significant, with so many unknowns and such high stakes.  

  2. I’ll spare you the politics of the answer to this question. 

  3. Including the 15-minute antigen-based test from Abbott that sounds like a game-changer. I figure there’s a long line of people waiting for access to those. That line of people probably includes Mark Cuban and Joe Biden, so I’ll take a number and wait. 

  4. Not to be confused with “RT-PCR,” which actually stands for “reverse transcription PCR.” Reverse transcription is the process of converting RNA to DNA, which can be done in a PCR (real-time or not), but the machine that does the reading is a “real-time PCR,” no abbreviation. You’d have thought there would be more distinct terms given how much linguistic space there is here. 

  5. Many years of scientific development have gone into making this possible, and the chemistry that makes this possible is both shockingly cool and shockingly routine. 

  6. The machine was from 2007, and allegedly in good condition. It was listed for $5,000, plus another thousand in shipping. We settled on $4,000 plus the thousand for shipping after the PCR failed one of the validation tests that it offered. (After contacting Applied Biosystems and talking to several helpful support people, I concluded that the “plate error” was likely spurious.)

    The package, shipped straight from Taiwan, required some extra justification when it hit customs at the Fedex distribution center in the US, but I detailed the purchase and they let it through. It took an hour to deconstruct the box and wooden frame the PCR was shipped in (it really felt like the $1,000 shipping charge was warranted), and even still, the case was clearly dented and misaligned, and the power cord was missing. The damage to the case proved only cosmetic, though, and I was able to power the machine up with a standard PC power supply. 

  7. More precisely, I only run tests for those people who I am willing to die for, who are unlikely to sue me, AND who don’t have easy access to fast testing. 


  1. This is so neat!

    Re: your relative risk calculations - my understanding's that the sensitivity of PCR tests doesn't reach 90% even in a lab where bare feet are forbidden - it peaks at ~80% when shedding is at maximum. :( Source:

  2. That’s a great point; one thing I didn’t go into detail on in this post is the timing of the test, which that article shows good numbers on. I try to wait at least four days after theoretical exposure to test, just to make sure whatever might be incubating has incubated. (This ends up being convenient, as I tend to test on Friday to protect the people I see that weekend from the people I saw the prior weekend.) The fact that you can shed viral particles without symptoms and before a PCR shows positive results is a good reminder that we have to be careful even with testing, and that education around good versus useless testing is really important.

  3. Daughters: Ooompa, Oooompa, Loompa Dee Doo

    Mom: Great work, Ooompa Looompas!

    Elon Must at his Gene Wilder house: I want my schtick back

  4. that is an impressive set up. Well done Karmel!


Post a Comment

Popular Posts