Menu Home

How does a COVID PCR test work? (part 4) – tying it all together

We’ve had quite a monster run on this COVID PCR test explanatory series.

Part 1 covered how to extract genetic material from your swab sample.

Part 2 covered the basics of a standard PCR.

Part 3 went over the very basics of an RT-PCR, which is what’s used for COVID PCR testing. We went over how it differed from a standard PCR, in that it uses:

  • RNA as its starting material
    • Meaning it needs a reverse transcriptase to convert the RNA into cDNA
    • Which means there’s an initial reverse transcription step in the run/thermocycling, and
  • It uses a fluorescent probe

If none of these terms are making sense, I would highly recommend reading part 3!

Here’s a visual representation of a typical RT-PCR cycle as a reminder.


Given we’ve already covered what goes into the individual tube for an RT-PCR run, today we’re going to focus more on the actual RT-PCR machine, or the thermocycler itself, and how to interpret the results.

CFX96 Touch Real-Time PCR Detection System | Bio-Rad Laboratories
QuantStudio™ 6 Flex Real-Time PCR System, 384-well, laptop
This one in particular is 1.1 m x 0.7 m x 0.9 m in dimension

RT-PCR machines tend to be significantly bigger than your ordinary PCR machine (although some can be quite compact) and are generally plugged into a PC, which has the program that lets you customise each run to suit your needs.

BioRad CFX96 Touch Real Time PCR - YouTube

The touch screen displays on the thermocycler itself are usually just for simple operations, like opening the lid, starting up the run, pausing it, or checking how long it has until it finishes. The actual set up is usually done on the PC application/program for that particular machine, which is usually provided by the manufacturing company at a cost.

This is in contrast to a standard PCR, which can be programmed and operated as a standalone unit. You don’t need to have it plugged into a PC for it to work.

So why all this complexity?

Well, it comes down to that fluorescent probe that’s in the reaction tube. You know, the one that emits light.

If you remember from part 3, every time the reaction goes into extension phase, a fluorescent reporter is released and able to fluoresce (emit light). This means that, the more copies of cDNA you make, the more light is being emitted. I.e. The number of copies is proportional to the amount of light being released.


But to explain fluorescent reporters, let’s take a step back to go over fluorescence in general. This is going back to high school physics.

Fluorescent things emit light.

Light doesn’t just magically appear, either. For a fluorescent object to fluoresce (i.e. produce light), it needs to be bombarded with light first. Then it refracts that light back out, and it looks like the object is lit up.

Take a sheet of white paper under UV light. It looks like the whole thing is glowing, because the white is bouncing back more of the light towards you than say, a black sheet of paper, which absorbs more light.

Get that white sheet of paper and write something with fluorescent highlighter. The writing will be even brighter again. Highlighters have been designed to bounce back more light (i.e. absorb less light) so that it looks brighter and stands out compared to ordinary ink.

What makes a "fluorescent" highlighter marker so bright? | Science  Questions with Surprising Answers

You can see the different wavelengths of light (i.e. colours) in the above diagram. That’s all it is. Colours are just light travelling at different wavelengths. Think back to all those boring diagrams in school science textbooks.


So, with that in mind, a fluorescent reporter has been designed to do something similar- except it’s far more precise in the light that it gets bombarded by (excitation wavelength), and the light it bounces back (emission wavelength).

As in, a specific fluorescent reporter requires exposure to light at precisely ‘x’ nanometres (‘nm’ – the unit of measurement for wavelengths), and it will emit, or release light at precisely ‘y’ nanometres.

Exposing it to normal, ambient light will still make a reporter emit light at y nanometres, because ambient light contains all spectrums of light (including light at x nanometres). But because it’ll be exposed to only a small amount of light at x nanometres, it’ll only release a small amount of light at y nanometres…

Hopefully that makes sense? If you don’t bombard a reporter with a lot of the light at the correct excitation wavelength, then it won’t release a lot of light at the specific emission wavelength. It’s proportional.

The more intensity at which you bombard a fluorescent reporter with the excitation wavelength it needs, the more intense emission you get.

But reporters don’t emit light forever. The more light it emits, the faster it’ll gas out and stop emitting all together.

So you have to find a happy medium when it comes to exciting reporters. Too much and you’ll lose signal too quickly. Too little and you won’t detect anything.


Okay- still with me?

So, the reason why the RT-PCR machine is so bulky is because it has the capacity to bombard your PCR tube with different excitation wavelengths, and it also detects the emission wavelength that comes out of that tube.

Fluorescence Microscopy- Definition, Principle, Parts, Uses

Within the RT-PCR machine, there’s a light source. It can emit the whole spectrum of visible light.

But then, there’s also a filter (excitation filter), which can literally filter out a very specific wavelength of light (blue in the above figure, but it could be any colour) at the optimum intensity. The sample tube with the reporter (specimen) is then exposed to this specific wavelength of light, and it releases its specific emission wavelength (green in the above figure). This then goes through a second filter (emission filter) and gets detected by the machine. Ignore the image thing- it’s irrelevant to this particular technique.

Each reporter will have its emission wavelength specified in its instruction manual (product sheet), so all you need to do is to make sure that you tell the RT-PCR program on the PC to excite the sample and look for light for that particular wavelength. You can do this by selecting the right channel.

You can see the list of channels in the row titled, ‘Reporter dye’, in the below figure. You have FAM, HEX, Texas Red, Cy5, and Quasar 705 as example reporters. These are real reporters, by the way.

If you select the FAM channel (Channel 1), for example, then the machine will automatically excite the samples at a wavelength between 250-490 nm and detect the resulting emissions between 510-530 nm, because that’s what you need to use for the FAM reporter. If you select the Cy5 channel (Channel 4), then the machine will automatically excite between 620-650 nm and detect emissions at 675-690 nm because that’s how the Cy5 reporter works.

Here’s a question to test your understanding.

If I had a reporter that emitted light at 630 nm, what channel would I choose on the RT-PCR for detection?

If I had a reporter that emitted light at 570 nm, what channel would I choose?

Have a think… It’s not a trick question…

Well, hopefully it wasn’t super complicated, but for a reporter with an emission wavelength of 630 nm, you’d choose the Texas Red channel (channel 3), because it detects between 610-650 nm.

For a reporter that emits at 570 nm, you’d use HEX (channel 2), because it detects between 560-580 nm.

The machine will automatically excite at the required wavelengths as soon as you select those channels, so you won’t have to worry about setting that part.


In addition to being able to bombard and detect specific wavelengths of light, the RT-PCR machine can also turn the amount of emitted light detected into a numerical value. Usually it’s called relative fluorescent units (RFU), but essentially, the more light is detected, the larger the RFU number.

And if you remember, I mentioned that the amount of reporters available to fluoresce and emit light is proportional to the number of cDNA copies produced. This means that as the reaction takes place, there should be an exponential increase in the amount of emitted light coming out of a reaction tube, because the cDNA copies increase exponentially.

And that’s exactly what you see in a RT-PCR run.

On the y (vertical) axis, you have the RFU. Again, the higher the RFU, the more light is being detected by your RT-PCR. The x (horizontal) axis shows you the number of PCR cycles that have happened. The first image is in log scale, but sometimes it’s easier to see all the results using a linear scale.

Now you can see all the lines more clearly.

Each of the thin, blue lines is an individual reaction tube. You can see that in this instance, some of the tested samples had some target RNA material to amplify (this example uses the E gene from SARS-CoV-2), while for other samples, there was no target RNA template, so the primers weren’t able to bind anything and no amplification occurred.

Because you can see the graph being generated as the reaction happens, this type of PCR is called a ‘real time’ PCR.

The other thing you’ll see is the thicker blue line that cuts across some of those exponential curves.

This thicker line is called the threshold, and it might need to be dragged and adjusted up and down manually.

The threshold is really important, because it basically tells the machine at what point in your PCR cycle a strong enough emission of light was detected. I’ll go into it in more detail below, but you just have to make sure that the threshold is set so that it is well and truly above any ‘noise‘.

What do I mean by noise? Well, you can actually see it at the start of the reaction.

It’s these little blips.

There’s no way that enough cDNA (and reporter) are present to release a real signal of light this early in the reaction- so this is like a false signal. And you can tell it’s noise, because you can see the real curves later- they’re much larger and pronounced.

This one particular noise peak is fairly mild, but sometimes it can spike quite a bit.

Sometimes it’s because of technical error on the user (bubbles while pipetting samples and reaction mix, ink/pen/writing on the tube, lid not sealed properly on tubes), sometimes it’s the dye (expired, old, freeze-thawed too many times), or it’s the machine (glitchy light filters, glitchy detectors). Either way, it’s just something that happens.

You just need to make sure that the threshold line is above these noise peaks. I’ll explain why that’s so important in a sec.


The graph is nice as a visual representation of the results, but for record keeping purposes, what we need are numerical values. They come out looking like this.

So what does that mean? Well, the column titled, ‘Well’ indicates where on the 96 well plate the sample tube was.

The column titled ‘Fluor’ indicates the reporter being detected, which in this case is FAM.

The ‘Target’ is the target RNA template you’re looking for (E-gene, from SARS-CoV-2), while the ‘Content’ in this instance is ‘unknown (Unkn)’ because I’ve marked them all just as plain test samples.

The ‘Sample’ is what’s in the tube. For that reaction I have 8 different test samples, alongside a no template control (NTC) that has no RNA material added (just enzyme free water), and a positive control (PC) that I know has the target RNA sequence added. This way I know what sample each of those tubes had.

The most important thing after that is the ‘Cq‘ value. Sometimes it’s interchangeably called ‘Ct‘. Personally I’m more used to the latter.

Either way, this denotes the cycle number at which the emission signal first crosses the threshold.

I.e. when the signal was strong enough to be a ‘real’ signal, as opposed to noise.

So you can imagine, if I had a lot of target RNA (i.e. lots of E gene from SARS-CoV-2) in my sample, then it wouldn’t take very long for the line to hit threshold, because there’s just more detached reporters releasing fluorescent signal sooner.

And that would mean that the Cq/Ct value would be smaller.

Whereas if I had very little target RNA (i.e. not as much E gene), then it would take longer for the line to hit threshold, because it takes longer for there to be enough detached reporters fluorescing for the signal to become strong.

This would mean the Cq/Ct value would be higher.

And if you had no target RNA (no E gene at all), then the line would never cross threshold, so the Cq/Ct value would be N/A.

So the smaller the Cq/Ct value, the more target there was in your original sample. The larger the Cq/Ct value, the less target material there was in your original sample.

If you equate it to an infection, assuming that the sample was taken in the exact same way, then if you have a low Cq/Ct value, then you had lots of virus (or viral material) in that individual, and if you had a high Cq/Ct value, then you had less virus (or viral material) in that individual.

I say viral material, because RT-PCR doesn’t tell the difference between active and inactive viruses- we’re just measuring the number of copies of RNA there might have been.

But going back to thresholds and why the threshold setting is really important:

If the threshold was below noise level-

See how the Cq/Ct values have changed?

As opposed to earlier, which was more correct.

All of a sudden, samples that were N/A now have Cq/Ct values. Even the samples that had Cq m/Ct values before have completely shifted. This is a major problem.

That’s why you have to make sure your threshold is above noise- so you don’t get inaccurate results.


So how does all of this apply to COVID? Well- your primers and probe kit should have come with instructions on how to interpret the results for testing, but usually it’s something like:

If your sample has a Cq/Ct value <36 (less than, but not equal to 36), then the result is positive. If your sample has a Cq/Ct of 36 or greater, than the signal is too weak for the result to be accurate. The closer your Cq/Ct value is to that cut off, the ‘weaker’ the signal.

Sometimes a result might not have a nice, strong exponential curve. It can have a diagonal slant.

Maybe like this.

Those are all ‘weaker’ positives. The above example at least has a Cq/Ct value under 32, so it’s still safe, but if it were closer to 36, then I would recommend repeating the test, either with the same samples or a fresh one depending on how confident I felt about the user technique and sample quality, or maybe by using a different target (i.e. different primers and probe kit).

So with this knowledge in mind, how would you interpret this result?

Firstly you need to check your controls.

The NTC (no template control) should have no Cq/Ct value, because there’s no target RNA to amplify.

The PC (positive control) should have a strong (not near the cut off of <36) Cq/Ct value, because I’ve intentionally added target RNA. I know it’s going to have a Cq/Ct value.

Hopefully you can see that both are correct for this particular run. The NTC has a Cq/Ct of ‘N/A’ and the PC has a Cq/Ct of ‘31.65’.

With that, I can finally go on to interpret my test results. I’ve writen out the Cq/Ct values in order

  • Sample 1 – 28.17
  • Sample 2 – 27.02
  • Sample 3 – 27.36
  • Sample 4 – N/A
  • Sample 5 – 27.05
  • Sample 6 – N/A
  • Sample 7 – 28.27
  • Sample 8 – 26.49

So, what do you think? Who (unfortunately) has COVID and who doesn’t?

Well, thankfully all the Cq/Ct values above are well and truly below the cut off of <36 cycles, so the interpretation shouldn’t be hard.

Sample 1 – 28.17 (positive)
Sample 2 – 27.02 (positive)
Sample 3 – 27.36 (positive)
Sample 4 – N/A (negative)
Sample 5 – 27.05 (positive)
Sample 6 – N/A (negative)
Sample 7 – 28.27 (positive)
Sample 8 – 26.49 (positive)

Hopefully that made sense!

Here’s another thought exercise.

What if the samples looked like this?

  • Sample 1 – 28.17
  • Sample 2 – 37.21
  • Sample 3 – 27.36
  • Sample 4 – N/A
  • Sample 5 – 18.52
  • Sample 6 – N/A
  • Sample 7 – 41.08
  • Sample 8 – 26.49

Have a think…

Hopefully you’ve figured it out, but here’s the answer below:

Sample 1 – 28.17 (positive)
Sample 2 – 37.21 (negative)
Sample 3 – 27.36 (positive)
Sample 4 – N/A (negative)
Sample 5 – 18.52 (positive)
Sample 6 – N/A (negative)
Sample 7 – 41.08 (negative)
Sample 8 – 26.49 (positive)

Because while samples 2 and 7 have a Cq/Ct value, it’s above the cut off of <36. This means that those samples are negative.


A good diagnostic lab will always keep the results of each run on file. It’ll usually have the following details on hand, bare minimum:

  • Run file name (the actual data from the PCR run that can be opened in the RT-PCR program on the connected PC)
  • Date
  • RT-PCR machine (i.e. which machine was used)
  • User (name of person who performed the run)
  • Target (what were you testing for)
  • Primer and probe kit used
    • Kit lot number
    • Expiry date
  • RT-PCR enzyme kit used
    • Kit lot number
    • Expiry date
  • The plate layout and what was in each well/tube (i.e. sample name, controls)
  • Results

All of these are useful for general record keeping, but also for when things go wrong. If results look funky and you notice it’s happening when the samples are run on a particular RT-PCR machine, then it might mean the machine is faulty.

If the results look funky when only person A uses it, then maybe they’re faulty, I mean- maybe they need to be re-trained.

If the results look funky and the same primer and probe kit were used each time, then maybe the primers and/or probe have degraded.

At least you can narrow down issues quickly.

On top of this, the Cq/Ct values of the positive controls are recorded for each run, too. This is important because you can check the value across multiple runs and users, and make sure it’s consistent. If there’s too much variation, then you run the risk of having the same sample come up as a strong positive and a weak positive, all because of that particular run (or user). You want consistency at all times.


And that concludes the ‘How does COVID PCR testing work?’ series! I hope you learnt some stuff.

Thank you to all those lab technicians who have been working tirelessly to process so many samples.

Thank you to all those workers who are collecting swabs and transporting them to testing facilities.

And thank you to everyone who has gone to get tested.

Stay safe!

🌻


Some other Q and As.


Why can’t we set up more COVID testing sites?

Well, the short answer is time, resources, and money. I’ve shown you how much stuff is needed to set up a basic lab in this four part series, but you have to remember that it has to be safe for the technician to do the tests. I’ve worked in outreach for about 18 months now, where the project has involved setting up shipping container labs to send to the Pacific. It comes with a BSCII and other lab equipment, as well as a PC and RT-PCR machine, and you essentially install all the benches and equipment on site. Obviously you need an external power source to run everything and keep the container cool, but it’s a massive undertaking.

But it can be done.

You also need to train people to do the technique. Here in Australia it might not be so bad, but in other countries you might be teaching people with basic high school science knowledge only.

It can be done, but it’s also a lot of time, resources, and money.

Mostly money, and they don’t just grow on trees, unfortunately.

I wish it did, though.


How do labs process thousands upon thousands of samples every single day?

Robots, for one thing.

I mentioned it in part 1, but you can get robots to do the entire extraction process (after adding the lysis buffer) for you. All you need to do is to provide all the buffers and it can do the rest. I think they can extract multiple 96 well plates at a time. Just depends on the robot.

To speed up the RT-PCR step, you can pool four or eight test samples in the one tube. That way you suddenly go from 384 samples (including two controls) to 1528 or 3056 samples (again, including two control wells). If a particular tube comes out positive, you can then split the four or eight samples into individual wells and test it again.

All of this requires the money and resources to have the equipment and reagents to do the testing.


What do you do with the RT-PCR tubes/plates after the run?

You take it out of the machine and chuck it straight into the bin.

You should never, ever, EVER, open a reaction tube after a run. EVER.

This is because you’ve essentially made more copies of genetic material than you had before, and you don’t want it getting out.

Not because it’ll harm you- genetic material in and of itself isn’t necessarily harmful, or capable of doing anything to us as humans.

It’s because it’ll contaminate the environment.

Again, not because it’ll do harm to the environment, per se.

Remember, this is a testing facility to detect genetic material from pathogens. You only want to find said genetic material in your test sample tubes only, not from the surfaces in the lab.

If your lab is contaminated, you run the risk of introducing the contaminant to a test sample that might otherwise have been negative for said contaminant.

E.g. If bits of SARS-CoV-2 E gene get into a patient sample that was otherwise negative, then that person’s test result is now going to be positive. Not because they caught COVID, but because you’ve contaminated their test sample.

And the annoying thing about RNA (and DNA) fragments is, while it degrades, it remains in the environment for a long time. The only way to clean it properly is to wipe everything down (meticulously) with a bleach solution.

Have you been enjoying wiping down every single surface to get rid of invisible germs? Well- same issue here. It’s tedious.

So the best way to avoid this is to just handle potential contaminants carefully and safely, and dispose them promptly after use.

Diagnostic labs will do regular swabbing of their lab environment to make sure that there aren’t any contaminants lurking on the surfaces. This is part of the regulatory requirements of running a diagnostic lab. They’ll keep records of these environmental testing PCR results so that if anyone needs to see it, they can.


If you’re setting the threshold manually, wouldn’t you get some variation between labs and people?

In short- yes.

But this is why you have to train people to be consistent. Hopefully the instructor/teacher will teach the trainee how to set thresholds, so that it’s the same across all members of the lab.

Sometimes we have panel samples that we test lab teams with. If the Cq/Ct values are too different from the main lab, then they have to train again so that it’s more consistent.

Setting thresholds is somewhat subjective, but it shouldn’t vary so much that a positive sample becomes negative and vice versa. That’s the most important thing.


What if I want to calculate how much virus was in my original sample?

One virus = one copy of the target sequence, right?

So if you can roughly calculate how many copies of a target sequence there was, then you can roughly equate that with how many viruses were present in your sample.

Again, bearing in mind that you can’t tell from an RT-PCR whether the virus was active or inactive (infectious or not infectious).

You can do this by doing a quantitative real time RT-PCR (RT-qPCR), or just qPCR for short.

This involves doing the exact same thing as usual, but alongside your test samples, you have to include known concentrations of genetic material, which are called standards.

So for instance, you can have the entire RNA genome strand from SARS-CoV-2, and have a reaction tube each that has:

100,000 (105), 1,000,000 (106), 10,000,000 (107), 100,000,000 (108), 1,000,000,000 (109) copies of SARS-CoV-2 genome, respectively.

You then run that alongside your test samples, and hopefully, your Cq/Ct values should reflect the 10 fold increase in concentration. This comes down to your pipetting technique (and pipette).

You can then plot that on a graph, and using a formula, work out how your Cq/Ct corresponds to how many copies of genetic material you may have had.

1.00E+05 28.475
1.00E+06 24.79
1.00E+07 21.47
1.00E+08 18.025
1.00E+09 14.8

At this point, you want the x axis scale to be logarithmic (otherwise the line will be curved), and the R2 value has to be as close to 1.0000 as possible. This tells you how ‘straight’ your line is. If the wonkier your line is, the less accurate the result will be.

And then you just solve for x by plugging in your corresponding Cq/Ct value (y).

Which I think is

e(Cq value-45.393)/(-1.482)

Using the above example.

Or if it’s excel, it’s

=EXP((Cq-45.393)/(-1.482))

Don’t worry, this is as complex as it got, in terms of maths. 😂

I have gotten an R2 value of 1.0000 before. I was so proud that I took a screenshot.

Again, I was very proud. 😂

Categories: General

Tagged as:

ABugsLife

A Ph. D graduate in Microbiology, residing in Victoria, Australia. Currently working in multiple locations but still in the STEM field. 👀 🦠 🧫 🧬

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: