As someone who did many RT-PCRs, the most widely-used method of detection for SARS-CoV-2 virus (the cause of Covid-19), during my PhD, here are my thoughts on why questioning the reliability of the PCR is just daft, but mostly it's just not the right question.

Thread: 1/n

To steal Ewan Birney's excellent analogy, your DNA is like a hard disk for long-term storage of instructions, RNA (spec. mRNA) is like RAM for short-lived, quick access to the instructions, and (my extension to the analogy) proteins are the operating system you interact with. 2/n

To make a protein (most of the "stuff" that makes up organisms), an instruction (gene) is read from DNA, into mRNA, and translated into protein.

So we use DNA to make the proteins we need, via RNA, and these are all specific* for a given species.

*to various degrees 3/n

PCR finds specific, unique sequences of DNA at either end of a region that you want to detect (often a part of a gene), makes a copy of the DNA sequence between, and does this loads of times until there enough to see (you *do* see it under UV light!) 4/n

The sequences at the ends of the region are called primers and are carefully designed to be specific for the thing you're detecting (remember I said that DNA was specific for each gene and also between species).

How do we know the original DNA sequence to make the primers?! 5/n

Well... we sequence whole genomes (the total set of instructions of an organism)!

This is done by chopping all the DNA up into little pieces, amplifying it by PCR, reading the sequence of those little bits, then piecing it back together like a huge, overlapping jigsaw. 6/n

This is like the Hayne's manual for your car (I'll be using more car mechanic analogies later!)

Ok, so we're amplifying bits of DNA that are specific to a particular gene or region and a particular species. THIS IS IMPORTANT! You're only ever amplifying DNA that is there. 7/n

This process works very well, has been done thousands of times a day, by thousands of scientists, over several decades, and has won a Nobel prize.

It's been used to detect inherited diseases, infections, new species, in drug design, to sequence genomes, EVERYWHERE! 8/n

SARS-CoV-2 is an RNA virus, so it's a bit weird in that its genome is made of single-stranded RNA, rather than double-stranded DNA like us.

PCR only detects DNA, but you can turn the viral RNA back into DNA with a process called reverse transcription (the RT in RT-PCR). 9/n

To see whether there's an DNA or RNA being detected, after PCR, you run the resulting mixture out on a gel, that uses the electrical charge of the DNA to pull it through a gel, bigger bits move slower, smaller bits faster, so they separate by size. 10/n

Then you add fluorescent dye that binds to the DNA in the gel, shine a UV light on it, and you literally see the DNA that's amplified.

You run a separate mixture of DNA fragments of known size and then you know the size of the amplified DNA too. 11/n

Running those gels takes time though (about an hour each) and manual work.

qRT-PCR adds the fluorescent dye right into the PCR mixture and then detects the amount of DNA amplified through time (a number of cycles of the PCR).

This means it's quicker and less subjective! 12/n

You get a plot like this, with amount of fluorescence (==DNA) on the y-axis and time (by number of cycles) on the x-axis.

Each cycle of a PCR amplifies each piece of DNA one time, and it increases exponentially (just like epidemic spread!), so you get tonnes quickly. 13/n

The Ct (cycle number threshold) is a threshold where you read off the amount of fluorescence, and if you do the same with say 10 solutions with known amounts of RNA and plot a standard curve, you can do a little maths and work out the amount of DNA in the sample. 14/n

You can do this without the standard curve, but then it's only semi-quantitative in that you don't know a definite amount of RNA detected.

BUT, you're only amplifying specific RNA (well, DNA) that was there in the first place (this needs repeating). 15/n

So where else could fluorescence that we detect come from?

1. Background fluorescence. You always run a water control to find background fluorescence that you can then subtract.

2. Contamination. Certainly! However... this is something that molecular biologists are... 16/n

...super-careful about and do everything to minimise. You use a new pipette tip with filter on (to prevent it getting into the actual pipette) for every sample. Use only sterile equipment. Keep everything super-clean. Wear gloves, lab coat, etc. And on and on. 17/n

Often this is learnt the hard way during your PhD when an experiment gets ruined through poor technique. You only do that so many times!

But contamination does happen. DNA and RNA is super-tiny and great at getting places you don't want it. 18/n

However, any contamination is likely to be at much, much lower levels than any RNA that's supposed to be there due to all the fastidiousness, and that will be reflected in how long it takes to amplify it (no. of cycles, remember the Ct?) 19/n

Could contamination come from anywhere else? Highly unlikely. Remember that we've designed the assay to detect RNA specific to the virus and can only amplify RNA that's actually there. 20/n

So with all this in mind, and given that we have tested thousands of samples to hone the technique, we can choose a Ct for our test that has an extremely good chance of giving a definitive yes/no to whether there's viral RNA present in a sample.

BUT...

21/n

...there are still gotchas! What happens if you have a borderline result where the amount of fluorescence takes just a little longer to appear? You retest.

Very important point: the Ct used is not always the same. It will vary depending on the machine used, kit used, ... 22/n

and more, but is always chosen based on real world data and sensible assumptions for the situation.

That's why it's ridiculous to ask what Ct a particular lab uses or to say that a Ct of X isn't suitable. It's meaningless, but has been seized on to rubbish the test. 23/n

It's the same as asking what tire pressure you use, then saying "Oh, that's not the right pressure, it should be X", or saying to a mechanic that they're using the wrong pressure. It depends on the car, conditions, etc. and you should assume the mechanic probably knows that. 24/n

Where else could problems crop up?

1. In the swabbing. The swab might not be taken from the right spot or not enough sample might be collected.

2. When the sample is being transported. Fortunately the swabs are specifically designed to preserve the RNA during transport. 25/n

Side-point: RNA degrades really quickly and easily, unlike DNA which can persist for many hundreds of years given the right conditions.

Remember, SARS-CoV-2 is an RNA virus. 26/n

3. Could we be detecting another seasonal coronavirus that causes cold-like symptoms instead? No, because our test specifically amplifies sequences unique to SARS-CoV-2.

4. Could we be detecting "dead virus" or "dead RNA"? Another side-point here... 27/n

...Using the word "dead" (search Twitter for "dead RNA" or "dead virus"! ) is really weird since i) by most usual definitions viruses aren't really "alive", and also can't really be "live" or "dead" since it's just a molecule. "Viable" is a better word. 28/n

It *could* be degraded though, i.e. broken up, fragmented. If there was degraded SARS-CoV-2 RNA present then chances are that i) it wouldn't float around for long (remember I said that RNA disappears quickly), and ii) it wouldn't amplify by PCR correctly. 29/n

Your body is an incredibly efficient machine. It doesn't keep stuff that shouldn't be there (redundant RNA from either yourself or anywhere else) around for long, and recycles and reuses that stuff quickly and efficiently. 30/n

Hopefully, all that has persuaded you that, first, the techniques used are i) reliable (decades of practice!), ii) based on many, many data points now, and refined all the time, iii) carried out by people who really know what they're doing.

Indeed, the testing... 31/n

...procedures are laid out in SOPs and covered by frameworks like Good Laboratory Practice (GLP) that mean that everything is highly standardised.

Second, this means the chances of false positives (calling a case positive when it isn't) are very small. 32/n

I'd be much more concerned about the possibility of false negatives, i.e. having something who thinks they've had a true negative result but are walking around while infectious.

Calling a false positive is still a concern because you don't want to isolate someone when... 33/n

...they needn't be, but it's erring on the side of caution, which is probably no bad thing during a pandemic that's killing hundreds of thousands.

So to sum up my Ted talk... 34/n

There are interesting and useful questions to be asked about testing accuracy, but saying:

1) that the "Ct for the test is wrong" is meaningless.

2) that "most cases are false positives" is incorrect. It's not possible. 35/n

3) that "dead virus" or "dead RNA" is being detected. Not really a thing.

4) that 5G is to blame and that vaccines will introduce malign nano-robots engineered by Bill Gates into our bodies. 36/n

Remarkably, all this stuff has been thought about long and hard by lots of scientists who actually know what they're talking about, and have no vested interest in getting it wrong. 37/n

Some much more legitimate questions to ask:

1) How can we further improve test reliability?

2) How can we reduce false negative rates that are a real and valid concern (though still not high)?

3) How can we improve test availability and self-isolation compliance? 38/n

Disclaimers:

I'm not a virologist, but I *am* a bacteriologist who studied a respiratory disease of cattle for his PhD.

I have no specific SARS-CoV-2 expertise.

I *have* done tonnes of RT-PCR, genome sequencing and assembly, and biological data analysis. 39/n

END.

Gah! Missed out "RNA" in 28/n.

It should have read: "by most usual definitions viruses aren't really "alive", and also **RNA** can't really be "live" or "dead" since it's just a molecule."

Also, thanks to @AdamRutherford, @ewanbirney, and troll-fighter @alanmcn1 for the inspiration to have a bit of a rant about PCR.