We return to the topic of Poisson statistics do delve a bit deeper into how digital PCR handles topics like false positives or ow abundance and rare targets. Dave Bauer is once again our statistics and math guru that brings his easy talking ways along with a depth of experience to give real-world examples that drive the concepts home.
The statistics of Poisson distributions can seem complex at first but are simpler than you think, which is important to know given their relevance to digital PCR. In short, they dictate the confidence you can have in the absolute quantification provided by dPCR.
Dr. Dave Bauer, Thermo Fisher Scientific’s very own dPCR Product Applications Specialist and statistics whiz, joins us once again for this short-but-sweet episode that’s a must hear for those working with rare, or low-abundance PCR targets. Dave and the hosts talk about applications like cancer research where these types of samples are common, and then get into the details of how the Applied Biosystems QuantStudio Absolute Q Digital PCR System works to provide elegantly simple technologies like false positive rejection, background subtraction, low dead volume microfluidic array plates, and a master mix with chamber loading dye. Join us to learn what each of these are all about and how they help to provide confidence and reliability in results that matter for your precious samples.
Visit the Absolute Gene-ius pageto learn more about the guests, the hosts, and the Applied Biosystems QuantStudio Absolute Q Digital PCR System.
Jordan Ruggieri00:00
Don't worry, Dave, in a world full of noise, you're that rare signal we're always searching for.
Dave Bauer, PhD00:06
Although someone's got to be the background subtraction.
Jordan Ruggieri00:08
So, I got, I got another one. You're a rare target, Dave, but we'll always make sure you're quantified to perfection.
Dave Bauer, PhD00:22
How about put me in the dead volume? How's that?
Jordan Ruggieri00:29
I came with puns. I came with puns, guys.
Christina Bouwens 00:31
You came with relevant puns. I just came with more like five-year-old jokes. Why is a koala not considered a bear?
Jordan Ruggieri00:38
I can't, I can't, already laughing. I don't even know.
Christina Bouwens 00:42
Doesn't have the right 'koala'fication.
Jordan Ruggieri01:00
Welcome to Absolute Gene-ius, a podcast series from Thermo Fisher Scientific. I'm Jordan Ruggeri.
Christina Bouwens 01:06
and I'm Christina Bouwens, and we're excited to be back in your feed this week with a new conversation with Thermo Fisher's own Dr. Dave Bauer, who you might remember from our awesome season one, episode eight, about Poisson distribution and statistics. If you haven't listened to that episode, we definitely recommend that you go back in your podcast feed and find it.
Jordan Ruggieri01:25
Today's chat with Dave is one in a series of shorter conversations that we plan to sprinkle in with our normal length episodes moving forward. These conversations will all be in-house with the amazing experts at Thermo Fisher Scientific and we're hoping that they'll provide some helpful education about the important aspects of working with digital PCR.
Christina Bouwens 01:45
So let's jump right into our newest conversation with Dave Bauer. Thanks for joining us, and we hope you learn something new.
Jordan Ruggieri01:56
Dave, it's great to have you here again on Absolute Gene-ius. We have the amazing privilege to have you as our first reoccurring guest. We're super excited that you're joining us again for this next season of Absolute Gene-ius. And maybe for anyone who hasn't listened to your previous episode, can you give us a little bit of background on who you are and what your position is at Thermo Fisher?
Dave Bauer, PhD02:20
Yeah, so I've been a Field Application specialist at Thermo Fisher for a couple of years, supporting our real time and digital PCR instruments. More recently this year, transitioned to a role that's focused specifically on digital PCR.
Jordan Ruggieri02:34
If you haven't listened to Dave's first episode on Poisson statistics and the math behind dPCR, it's amazing and really, really educational. We encourage you to listen to it, and Dave, you just have to remind us again, how much do you love the math for digital PCR?
Dave Bauer, PhD02:52
You know, I love it. It's just that it's not as complicated as people might think. Honestly, it's like four terms in the equation, but that's counting parentheses. It's a very simple thing.
Christina Bouwens 03:04
That's actually why I'm really excited about this. I'm going to call it a mini-series, but it's a really I want to get us to all together and create kind of these, these short, sweet episodes to dive a little bit deeper into some of the, you know, like on the more nerdy aspects of why digital PCR is so cool, but why it stands out. Because I think it's a really nice, approachable way to get your hands around digital PCR, whether you're really new to it or as just a basic refresher. So yeah, really excited to have you back on the podcast to just talk about some really cool topics.
Jordan Ruggieri03:32
Dave, maybe can you even just give a brief overview of Poisson statistics and maybe why they matter for digital PCR as a little review of your prior episode.
Dave Bauer, PhD03:43
Yeah, the main reason we use Poisson statistics at all is because, for the most part, we're not, quote, unquote, counting molecules. If we could just look in there and say, "Hey, of my 20,000 micro chambers, 400 were lit up. I've got 400 targets." Unfortunately, the way things move around and fill those chambers, you might have some that have multiple targets, so you need to account for that, and that's where Poisson statistics come in.
Christina Bouwens 04:09
So in your role currently, you're an application specialist, so you're seeing all the ins and outs of what people are doing every day with digital PCR, and we have people doing a lot of different things. And I think a really nice kind of transition from your previous talk, where we're talking about Poisson statistics, and you know, we often really think about Poisson statistics zooming in to like those rare targets and rare molecules. And there are a lot of things that come up when we think about rare targets and rare molecules, but can you kind of talk about some of the applications that you come across when people are looking for rare targets in dPCR?
Dave Bauer, PhD04:41
So I always like to define rare targets as well, it can be helpful. So things can be rare in the sense that they're just low abundance. So if I was joining qPCR and I had a really high CT, and then rare target sometimes, we mean it's there might be a decent quantity, but it's mixed in a background of a high abundance similar sequence. And that's very commonly investigated with cancer research, where maybe you have a specific base pair change that can be a marker for that cancer, but there's tons of the healthy sequence around it, so it's very rare in that sense.
Jordan Ruggieri05:16
Is there a situation when you're looking for a rare target, where, if they're looking to count, you know, say, two molecules and maybe there's a false positive, right. You're looking at something that's really rare. Can we dive into some of the features of the Absolute Q itself that allow for some of the false positive rejection and add that confidence that what you're actually looking at is true, is a real count.
Dave Bauer, PhD05:43
So for the Absolute Q system, the plate, it's not moving around during that whole PCR amplification. So it gives the system the ability to take more than one image. So it's actually taking an image early on. That's the background. It then remembers if there happens to be some dust somewhere. All that information gets collected and held onto. Then after PCR cycling, there's our endpoint measurement. We subtract the background out, that way it helps remove potential false positives from artifacts and dust and things like that. And then the second half, after background subtraction, there's the analysis of that individual micro chamber where the system can look at the distribution of pixels inside of that one micro chamber of the 20,000 for that one sample, and figure out, does that signal represent real PCR or is it some remaining artifactual signal to get removed?
Christina Bouwens 06:34
We're talking about rare targets. We're talking about low abundance. It kind of brings to my mind this idea that what people are actually looking for is confidence. Confidence that those rare targets are there, and that, you know, you can trust the data behind it. And it kind of goes back to, you know, our discussion about Poisson statistics, you know. How can we trust that, you know, the count is accurate? So can you talk about some of the different ways that digital PCR systems help to improve our confidence that what we're counting is accurate, that it's there and that we know, you know, we know the truth behind, behind, what the data says?
Dave Bauer, PhD07:05
Yeah, and I'll, you know, use if you have a low abundance of that target, you really want to know if I see something that looks like signal, is it real? Because if you're only hunting for a couple molecules of your target, seeing one or two spots light up, that could be fairly informative for you. Versus if you are expecting 1,000 and you got 1,002 it's a pretty minor difference, but if you're expecting it to be low abundance, you need to have that high confidence to say that's a real indication that my target is truly present. There's a couple ways that we handle that. I know there's false positive rejection. The idea of the system doesn't just measure the intensity of the signal and say, "Oh, it's high, therefore it must be positive." It's actually going so much deeper than that looking at the image, the image, the pixels themselves in the picture that makes up that micro chamber, saying, "Does that signal look like PCR amplification, or is it bright because of some other artifact? " That's a really important way to exclude those potential false positives.
Christina Bouwens 08:05
I know, like, just looking at the pixels, you can kind of tell, you might be able to tell whether that is, like a true PCR based signal or just an artifact. Since the plate isn't moving over time, we talk a lot about background subtraction. Is that another way that you can kind of remove artifacts from these reactions too?
Dave Bauer, PhD08:22
Yeah, absolutely. Because what I mentioned was kind of the kind of the end point where you're saying, "I have signal. Let me evaluate it." But before that, there's another layer of protection to say, "Why don't we just take an image of what the background signal looks like?" Yeah, there's some bright spots, maybe from some dust or whatever, or the signal in the sample intrinsically fluorescing. We can capture that before PCR amplification, and then just simply remove it and. Like you said, the plates not moving. So we're doing that every single run, every single sample, getting on its own personal subtraction.
Christina Bouwens 08:54
Dave, when you and I were talking last week, you actually brought up a really good point that I'd love for you to kind of talk a little bit more about in this idea that background subtraction can actually help with some of the baseline setting and understanding what true positivity or negativity would look like in a reaction?
Dave Bauer, PhD09:10
Yeah, so subtracting the background it can be really helpful for avoiding false positives, but as you're mentioning, it has another benefit separate from that. It just makes data analysis simpler, more convenient, more robust, because that background subtraction, it's normalizing things. And I'll borrow the analogy one of my coworkers always brings up about real time PCR, when you do real time PCR, there is a background subtraction for the exact same reason. You need to normalize the signal, it makes for more robust analysis. So by having our system take that background image. It also kind of sets the scale and level sets things so that we can have an easier comparison of sample-to-sample, assay-to-assay, condition-to-condition, rather than having, you know, there's the positive and the negative cluster. Without background subtraction, both of those might be moving up or down together and that makes setting the threshold more difficult. Versus when you subtract the background, you're kind of anchoring to one consistent spot for your analysis between comparisons.
Jordan Ruggieri10:10
When you're looking for targets that are so rare, what are the chances that you might actually have false positives? Like, are you if you're looking for one or two or five, you know, a handful of these, of these mutations, or these events, or these molecules, you know, is there when you, when you, actually run that through the system, is there, is there a chance that, that something might come up, that's a, that's a false positive, and does that impact the results that you're actually looking to get?
Dave Bauer, PhD10:39
So something to keep in mind, with rare targets is, by definition, there's not a lot of them, usually, and so you want to make sure they're getting into the analysis. So the key word in digital PCR around that is dead volume. So usually, unlike qPCR, you put your reaction mix into the well, all of it is contributing to the signal. But digital PCR is different, where we have to usually do something with that reaction mix to make our individual micro-reactions, and in that process, we can lose some of that reaction mix. But imagine if you only had one, two or three of your targets in that volume, and now you lose some of that volume. Well now you might be losing some of those rare targets. So you had three, and now you had one, or you had two, and now you have zero, and now you're not going to get a good understanding of what's truly in your sample. And that could be really important for certain applications, some more than others. One example is if you're doing single cell analysis, where you're actually lysing open that cell, then you want to get an understanding of what's inside of it, and so you're doing digital PCR with that. In that case, there's not a lot of targets. You might have a very low number of the sequence you're looking for, and you don't want to be losing it in the process.
Jordan Ruggieri11:47
Going really nerdy here and homing in on the Absolute Q. What are some of those high-level features and high-level ways that the instrument and the software and the consumables all work together to really help enable confidence that the quantification that's happening is accurate?
Dave Bauer, PhD12:09
I would divide it up into like three different things. One is the microfluidic array plate that helps reduce the dead volume in its design and it also contributes to the consistency. There's 20,480 micro-chambers per array, and we very consistently see 20,000 plus, very little variance around that. And that also contributes to low debt volume. And there's the master mix, what we're putting inside the plate. That has a special dye in there that's used as a reference for chamber loading. That way we don't accidentally have false negatives interfering with our results. And then honestly the system itself the Absolute Q with its image analysis, the way we take those images on those stationary plates, those stationary micro-chambers, we can then do a lot more algorithmic analysis on that.
Christina Bouwens 12:58
So when we take a look at digital PCR systems where you can, you know, have a lot of control over that data, I know that there's a lot on the market that give the users just an incredible amount of control over what's a good versus bad data point. Can you talk a little bit more about that?
Dave Bauer, PhD13:14
Yeah. In science, one of the most critical things is objectivity. We want to eliminate the user making decisions as much as possible, because then you're not going to have reproducible results if analyst A is doing something different from analyst B. So the Absolute Q system with that false positive rejection, it's automated. The system is doing it the same way every day, regardless of who hits the start button, versus requiring your analyst, who has a various degree of training and making them apply a subjective decision. And that can dramatically change the results, especially if you have rare target stuff, if they identify one of the two real events as negative, now you completely change that answer.
Christina Bouwens 13:56
What kind of tools are we using to kind of support that confidence, like, what layers of confidence do we have? You say we have automated tools to support that rejection or acceptance. Can you talk a little bit more about that?
Dave Bauer, PhD14:08
Yeah, so the system has the algorithm based on AI, where it's learned from true positives, it's then learned what that looks like in the signal of those individual micro-chambers. Because you can imagine, with 20,000 plus micro-chambers, you're not going to want to sift through all of those data points. The system is doing it for you and then just informing you and it's also keeping kind of a record of it. If in the software, it shows you those rejected spots, you can see if there's some kind of unusual trend. The user has some awareness of that, it's not just a metaphorical black box.
Jordan Ruggieri14:44
Dave, you're out there. You're talking with customers. You're talking with digital PCR users all the time. If somebody is considering bringing digital PCR into their laboratory, what should they be looking for? What are questions they should be asking themselves? What are things that they should keep in mind as they're evaluating digital PCR platforms?
Dave Bauer, PhD15:01
Yeah. So digital PCR a lot of different applications, a lot of reasons to use it. And the Absolute Q checks a lot of those boxes, for me at least as a scientist. We know that with digital PCR, you might have rare events, whether it's low abundance or it's rare because it's one of the few wild, few mutants within the wild type population. In either case, you care about those rare positives, and the Absolute Q has a lot of features, the background subtraction, false positive rejection, so you have the confidence so that you're detecting true signal on those rare events. That's always important. And then sometimes precision, sometimes you're doing digital PCR for precision. And a lot of precision coming from Poisson statistics has to do with the number of micro-chambers you're using. And if your system has variable numbers of micro-chambers or micro-reactions run to run, that impacts the precision. It makes it more difficult to compare results across runs. Another critical benefit of digital PCR is the absolute quantification. So a lot of people doing real time PCR might say, "Hey, I get quote, unquote, absolute quantification," but that's always relative to a standard. There's always got to be a standard that you're comparing your CT values in that case. And there are a lot of potential issues dealing with standard curves. One of the biggest is what's your reference material? And if you're making it yourself, you obviously have a way to quantify it very accurately, so why are you using real time PCR anyway? And if you're getting it from a vendor, you don't always know how they're quantifying and how accurate their results are. And I can't tell you the number of times I've seen people switch digital PCR and say, "I'm getting different absolute quantifications from real time." And I say,"Run that standard if it says it's 10,000 copies, put it in the digital PCR system and see what it says." And it isn't always what it says,
Jordan Ruggieri16:46
Dave, you're one in a million, but dPCR would find you in a heartbeat.
Christina Bouwens 16:59
That was Dr. Dave Bauer, Application Scientist at Thermo Fisher Scientific. Thanks so much for joining us for today's episode of Absolute Gene-ius. It was produced by Matt Ferris, Sarah Briganti and Matthew Stock. Stay curious, and we'll see you next time.
Christina Bouwens 17:13
I had a whole sticky note just for jokes. And this is another Olivia classic, ready? She dropped this on me. What do you do if you're addicted to seaweed?
Jordan Ruggieri17:23
What do you do if you're addicted to seaweed?
Christina Bouwens 17:27
Sea kelp.
Jordan Ruggieri 17:30
Sea kelp.
Christina Bouwens 17:34
It was so cute. I was like, that's adorable.