AVENIO Tumor Tissue CGP Automated Kit

In the pursuit of precision medicine, every genomic insight is crucial. Backed by the trusted expertise and proven technology of Roche and Foundation Medicine. The AVENIO Tumour Tissue CGP Automated Kit helps you bring comprehensive genomic profiling research capabilities in-house with ease.

Designed to empower your laboratory with a complete end-to-end five-day workflow from tissue DNA extraction to result generation, with ready-to-use barcoded reagents, robust bioinformatics, and true walk-away automation with the AVENIO Edge system.  

Leverage a fully automated system for library preparation, target enrichment, on-deck pooling, normalization, and quantification. AVENIO Tumour Tissue CGP automated kit is designed to offer simplified batching flexibility, adapting to your labs unique operational needs.

In a single workflow, discover insights from four genomic alterations and signatures: TMB, MSI, gLOH and a pan-tumor HRD biomarker.

Experience excellent performance and confidence with our advanced genomic analysis technology with the FoundationOne® Analysis Platform, a cloud-based secondary analysis software built on Foundation Medicine’s, proprietary algorithms and insights from high-quality genomic profiles. Unlock the power of insight excellence with the AVENIO Tumour Tissue CGP Automated Kit to enable your lab to lead the way to groundbreaking discoveries in advancing precision medicine.

Kira Curtin: Welcome, everyone. I'm Kira Curtin, senior editor at GenomeWeb, and I'll be your moderator today. The title of today's webinar is ‘Automating Comprehensive Genomic Profiling, a Global Multi-Site Reproducibility Study’, and is sponsored by Roche. Our speakers today are David Gonzalez, a professor at the School of Medicine, Dentistry, and Biomedical Sciences at Queen's University of Belfast. Sunandini Chopra, Senior International Product Manager, Oncology and Genetics at Roche Diagnostics. And Tara Love, Global Medical Affairs Lead, Precision Oncology and Informatics at Roche Diagnostics. You may type in a question at any time during the webinar. You can do this through the Q&A panel, which appears on the right side of the webinar presentation. And then if you look to the bottom tray of your window, there are a series of widgets to enhance your webinar experience. And with that, I'll turn it over to David. David.

David Gonzalez de Castro, PhD: Thank you very much, Kira. So welcome, everyone, to this webinar. We're going to tell you about new, exciting developments in comprehensive genomic profiling and the value of automation. 

So I'm just going to do a quick introduction just to set up the scene very briefly. So basically, I guess if you've joined this webinar, I guess you probably agree with me and my colleagues that comprehensive genomic profiling is really where the money is. It's really what we need to do for cancer patients. And when we talk about comprehensive genomic profiling, we mean just going beyond the sort of handful of hotspot mutation testing. and just going into a test that as a single tool can actually identify and detect not only sequence variants in the most common actionable genes, but in a large series of genes involving cancer, but also, very importantly, different alterations that are equally actionable and critical for the biology of cancer, which are copy number alterations, gene rearrangements, and fusion genes, and also complex signatures, genomic signatures, such as tumor mutation burden, microsatellite instability, loss of heterozygosity, and homologous recombination deficiency. 

So a test that detects all of these different alterations is what we really need for successful sorts of cancer genomic profiling. And we can also see how the clinical trials and the approvals of different drugs have been focusing on targeted therapies relating to genomic alterations. Even despite the fantastic improvements that, for example, immunotherapy has been bringing to cancer patients over the last decade, we still see an enormous amount of trials that are driven by the presence of driver alterations that potentially can respond to specific targeted therapies, for those patients with those alterations. So it's critical that we detect these alterations in the genome. 

Unfortunately, trying to get this level of comprehensive genomic profiling into our labs is not an easy task. It could be very overwhelming, particularly when most of these assays are manual, which are error-prone and potentially complex for staff in the lab that are used to different technologies. Depending on the size of your lab and the support that you have from different sorts of scientific and clinical input, it can be more complex than is necessary. It's also the fact that sometimes bioinformatics analysis could be quite intensive, particularly for tertiary analysis. And of course, as usual, we all rely on our excellent staff members to be able to just try and bring these assays into the practice in our labs. And of course, we're very busy labs. We really struggle to keep things going on a daily basis. So bringing new things are a quite significant challenge. 

So in an ideal world, what we really want is a user-friendly workflow that basically gives us complete accuracy. And what I mean is being able to trust any alterations that have been detected and reported completely without any shadow of doubt. It needs to be fully automated, not just part of the workflow, but the whole workflow needs to be fully automated to reduce and completely minimize the human intervention and potential errors. And also, it needs to be streamlined. We don't need our guys in the labs to just having to jump through hoops to go from one part to another part and just become very complicated to bring on to real life practice. So we need something that covers all of them. And I think that's what we're here today to hear about. And I'm going to hand over to our colleague in Roche, Sunandini, who's going to introduce the Avenio tumor tissue CGP automated kit. Sunandini, please.

Sunandini Chopra, PhD: Thank you, David. So hi, everyone. My name is Sunandini Chopra. I'm a Senior International Product Manager at Roche Diagnostics. So before we dive into the assay performance and study results, I'd like to walk you through some of the products and components that were part of this study. Here are some disclaimers. 

So I want to start by highlighting Roche's strong commitment to advancing comprehensive genomic profiling. So our portfolio includes both send-out and in-house solutions to support CGP testing, such as the Foundation Medicine send-out assays offering high-quality CGP testing, as well as the Avenio CGP kits that are in-house CGP kits supporting both manual as well as automated workflows. enabling labs to perform CGP testing on-site. We also provide a range of instruments and software solutions that support the entire CGP workflow and have been used to generate the data that will be presented later today. 

So these include the Avenio Edge Liquid Handling System that enables easy and turnkey automation of the NGS workflow, improving efficiency and consistency. The FoundationOne analysis platform that supports robust secondary data analysis enabled by Foundation Medicine. The Avenio Connect software that streamlines data transfer and integration within the laboratory ecosystem. And finally, navify Mutation Profiler that supports tertiary analysis and report generation. I want to highlight that the Foundation One Analysis Platform and the Avenio Connect software are only available with the Avenio CGP kits. So together, these components truly empower global access to CGP by offering a scalable, flexible, and integrated solution when put together. 

However, today's presentation focuses on the Avenio tumor tissue CGP automated kits. They are Roche's in-house comprehensive genomic profiling solution designed to enable decentralized CGP testing globally. So these kits are the result of a partnership between Roche Diagnostics and Foundation Medicine, more accessible worldwide. So this partnership brings together truly the strengths of both organizations. Roche contributes robust NGS workflows, an extensive global footprint, and a strong service and support infrastructure. Whereas Foundation Medicine offers over a decade of CGP leadership, demonstrated by delivery of more than 1.3 million CGP reports, publication of thousands of peer-reviewed manuscripts, and holding the largest number of companion diagnostic claims in oncology. 

So together, we offer to our users a CGP assay that features a panel design aligned with the FoundationOne CDX panel design and bioinformatics powered by the FoundationOne analysis platform. 

Another key component of the Avenio Tumor Tissue CGP Automated Kit workflow is the Avenio Edge Liquid Handling System. So I'm going to take a couple of minutes to walk you through this. So this system delivers simple turnkey automation for critical steps in the NGS workflow that includes DNA bead cleanup, library preparation, target enrichment, on-deck quantification, normalization, as well as pooling. So the Avenio Edge system simplifies the process for our users in three easy steps. So first, create the Avenio Edge work order file. Second, load the instrument, and 3rd and final, simply walk away, enabling hands-free automation of essential and complex NGS steps. So this streamlined automation improves consistency, reduces hands-on time, and really helps labs scale CGP testing with greater ease and confidence. So together, these components enable the Avenio CGP automated Kit workflow to offer a streamlined workflow that takes less than five days with only three and a half hours of hands-on time. 

So all the reagents required to run this entire CGP workflow, including the DNA extraction kits, are included in the Avenio Tumor Tissue CGP automated kit, obviously, except the sequencing reagents. So in this workflow, we start with a flexible tissue digestion and partly automated DNA extraction step. This is followed by a fully automated library preparation and target enrichment enabled by the Avenio Edge system that allows users to walk away for 24 hours while processing 12 to 24 sequencing-ready libraries at once. while using a 335 gene panel that is aligned with the FoundationOne CDX panel design for enrichment. 

The workflow supports flexible batching as well, with throughputs of 12, 24, and 48 samples, depending on the sequencer used. So currently, this assay is compatible with both the Illumina NextSeq 500-550, as well as the NovaSeq 6000 to meet varying user throughput needs. As I mentioned earlier, for secondary analysis, it utilizes the Foundation Medicine's bioinformatics pipeline to provide high-quality genomic insights. And lastly, this workflow is optimized for compatibility with navify Mutation Profiler for tertiary analysis and report generation. So with that overview, I will hand it over now to Dr. Tara Love to walk us through the assay performance.

Tara Love, PhD, CGMBS: Thank you so much, Sunandani. Hello, everyone. I'm Tara Love, and I'm the Global Medical Affairs Lead of Precision Oncology and Informatics at Roche Diagnostics. And I'm really happy to be here today to talk you through the performance results for the Avenio Tumor Tissue CGP Automated Kit. Now, what I'm going to be doing is taking you through a set of assay performance results that were generated by testing a total sample set of nearly 6,000 samples. And I want to highlight that the data sources and the data that I'm presenting to you today actually come from several studies outlined on this slide that were presented at the Association for Molecular Pathology Conference late last year in Vancouver, British Columbia. And various members of the Roche research and development team presented these data.

So let me start by setting the stage a bit. Assay performance was actually measured on a total of 5,844 samples. And that's a very large sampling to really allow for some robust measurements. And within this set of samples, there were just over 1,000 unique tissues. And that amounted to 40 different tumor types. And these could be further broken down into 120 disease ontologies. In terms of logistics, there were actually a total of 255 Avenio Edge runs, and those amounted to 140 sequencing runs on the NovaSeq 6000 platform. And then just around a little bit of testing design details, there were a number of operators was 13, and there were three reagent lots and two different testing sites. And just to make it clear, it was quite unbiased when we were looking at accuracy, where reference data were generated externally, on which we compared the Avenio Tumor Tissue CGP Kit. 

So in terms of, let's dive right in, and I'd like to first start by highlighting data on the end-to-end sample pass rate. And so to look at the sample pass rate, our R&D team analyzed 1,218 FFPE samples, representing 609 tissues. So these were run in duplicate. And this was 37 different tumor types. And you can see this distribution here on the left-hand side, where the most highly represented cancers were colorectal, breast, non-small cell lung, ovarian cancers, and melanoma.

You can see on the right-hand side that we observed an initial attempt pass rate, end-to-end, of 93.8%. And then upon retesting, an overall sample pass rate of 99.3%. And the team even stratified the pass rates by different tissue types, and they did not find any significant differences attributed to specific tissues. And it's important to note here also that only two samples out of this set of the 1,218 samples failed sequencing QC metrics. And importantly, even factors like necrosis levels ranging from 0% to 70%, and also melanin and melanoma samples, these factors did not have any effect on key quality metrics. 

All right, so let's talk a bit about pre-analytical requirements for successful CGP. So it's important to point out that the minimum thresholds for inclusion of samples were set at 20% tumor content, and a 10 square millimeter tissue surface area, and a minimum DNA input of 40 nanograms. And so the majority of these samples could actually be classified as having greater than 10 square millimeters tissue surface area. And in that, you know, in those groups of samples, 99% of these samples met the 40-nanogram minimum DNA input, while 90% actually achieved the optimal 300-nanogram DNA input. 

And I want to point out that even below the 10 millimeter, square millimeter surface area, the team looked at these 12 samples that had very small amounts of tissue, and 100% of them actually obtained the minimum DNA input level. And even 58% of those at those low levels achieved the optimal DNA input level of 300 nanograms. And in general, what you're seeing, you know, across the range of tissue surface areas is that you can see 99% to 100% of samples achieving minimum DNA input levels, and then 90% to 99% achieving the optimal DNA input levels. So these data suggest that our DNA extraction procedure is very robust across varying tissue surface areas. 

Okay, so let's now dig into how well the low DNA input samples performed in terms of quality control metrics and variant calling. So I just mentioned that the recommended DNA input for optimal performance is 300 nanograms. And so the Avenio assay was actually evaluated across a broad range. You can see from 10 nanograms to 600 nanograms. And then over here, you see various QC metrics. And really, the vast majority, and here on the left is 10 nanograms, the vast majority achieved good coverage at greater than 100x. And you can also see median coverage still being strong at 10 nanograms, around 500 reads. And so you know, you can see a variant call rate also of 99.3% at 10 nanograms of input over here. And then that starts to approach 100% with 20 nanograms of DNA input. And still we're well below the minimum, you know, required input of 40 nanograms. And so this, you know, really in the real world, we're dealing with situations where tissue quantity or DNA yield may be limited. And so this highlights the assay's adaptability in such real-world conditions. 

Okay, so now I want to take you through the DNA extraction performance. So what we did was we compared the Avenio Tumor Tissue CGP Automated DNA extraction kit against two commercially available methods, the MagBind FFPE DNA kit and the Qiagen AllPrep DNA/RNA FFPE kit. And so in a side-by-side comparison using 420 FFPE samples, Avenio consistently delivered higher mean yields. And you see this by the scatter plot here, where Avenio is plotted on the y-axis, and the competitor or the other method is on the x-axis.

And this also beared true with the QIAGEN All-PrEP kit as well, where you see a higher mean DNA yield happening with the Avenio CGP extraction. And then quality was looked at of the extracted DNA for the QIAGEN All-Prep comparison. And what we learned was the Avenio kit was comparable to that, the Q scores, the quality scores obtained with the QIAGEN kit. So what do we see here in general? We see higher yield and comparable Q scores. And it's really indicating that the Avenio extraction workflow is not only efficient at maximizing DNA recovery from FFPE samples, but also the integrity of the DNA is maintained. And we all know that is quite essential for doing optimal downstream NGS applications. 

So now let's dig in a little bit to sequencing coverage at various input amounts. So what we find here is sequencing performance was actually evaluated across a total of over 3,000 FFPE tumor tissue DNA samples. And these ranged from 13 nanograms to 300 nanograms input DNA. And the key metrics that were measured here were median unique exon coverage and the percentage of exons that are covered at greater than 100x. So across the range of inputs, you could really see the assay demonstrated consistently high coverage here in this column. And then even at 13 to 40 nanograms-- again, that's below our minimum required input level-- there was high coverage of 1,218 reads. And just to point that out, because in the real world, we do have these cases where you may have less. This highlights that the assay is strong in coverage at lower DNA input levels. So let's now delve into the sensitivity, or we can refer to this as the limit of detection. 

Now, this was assessed using a panel of 34 FFPE-derived DNA sample pairs. And that was across 12 tissue types. And so 13, you see on the slide here, 13 key variants, they span actually 4 variant classes, single nucleotide variants, indels, copy number alterations, and rearrangements. And these were analyzed at 5 different titration levels. So, at a minimum DNA input of 40 nanograms, The Avenio assay demonstrated a variant-specific limit of detection between 1% and 3%, 3.1% to be exact. And this is for key known oncogenic driver variants in genes like EGFR, PIK3CA, BRAF, and the RAS genes. And so just in general, for platform-wide limit of detection, this was achieved for single-nucleotide variants at 0.9% to 16.6%, and for indels, 1% to 14.4%. And so for the indels, these were indels in difficult regions, like homopolymer regions and cyanine alu regions. So in general, we're really seeing high analytical sensitivity with this assay. 

So, when you're looking at sensitivity, you should also be looking at specificity. And so the team did that by assessing the limit of blank. And this study for specificity was conducted across 69 non-cancerous FFPE tissue samples across 15 tissue types. And, you know, each sample was processed at a minimum of 40 nanograms and standard 300 nanogram input levels. And this was done in triplicate, actually, and it, you know, resulted in 414 assessments. And so the good news is the rate of false positive variant calls was extremely low across these assessments at less than 0.007% all the way down to zero for the various biomarker types, as you can see here. So we're seeing an analytical specificity of 99.99 to 100% for this assay across these different variant types. 

So a concordance analysis is what I want to talk about on this slide. It was done to look at accuracy against a reference method, so against an independent alternative CGP reference method. And so 856 FFPE tumor specimens were looked at, representing 38 tumor types. And the analysis focused on 324 overlapping genes. So for known pathogenic single-nucleotide variants and indels, the assay achieved a positive percent agreement, or a PPA, of greater than 99%. And for copy number alterations and rearrangements, the PPA was greater than 94%. You can see that here. These findings demonstrate strong alignment with an established reference standard, and it validates the assay's accuracy across this spectrum of genomic alterations. 

So let's zero in on various oncogenic driver biomarkers in the different alteration categories. So there was a very nice demonstration of agreement between a venial tumor tissue CGP automated assay and an independent alternative CGP reference method. And the punchline is really that we saw overall greater than 90% positive agreement in detection of key biomarkers. And in many cases, 100% agreement, as you can see here in this column. So we're highlighting here the ability of the Avenio assay to enable labs to do accurate identification of both established and emerging biomarkers across a diverse set of tumor types. 

So we also wanted to look at genomic signatures, complex signatures, such as TMB score, MSI score, genomic loss of heterozygosity score, and our HRD-SIG score. And we wanted to look at accuracy against a known reference method. And what we see in general here is for these complex genomic signatures, there was an analytical sensitivity of more than 91%. and an analytical specificity of greater than 98%. 

So that's really the net of all. And I've just given you all our performance results that were presented last year by our R&D team. And I want to thank you kindly for your attention as I walked you through this. And without further ado, I'd like to turn your attention to Professor David Gonzalez de Castro, as he's going to tell us a little bit more about Avenio Tumor Tissue CGP Kit in terms of a multi-site reproducibility study conducted across four independent sites. Thank you.

David Gonzalez de Castro, PhD: Thank you very much, Tara. So it's my pleasure to be here today presenting you the results from our multi-site study on behalf of my colleagues and co-authors in the other labs. First of all, I just want to give a brief introduction of our setup in Queen's University, Belfast. So we basically work in the Precision Medicine Center, which is a clinical laboratory with ISO-15189 accreditation for cancer genomics, bioinformatics, tissue hybridization, and digital pathology. And we provide comprehensive genomic profiling for both solid tumors and hematological malignancies. 

We were very, very lucky to have the first Avenio Edge System installed outside of Roche Laboratories back in February 2022, thanks to a collaboration that we were undergoing at the time with Roche. And we basically performed around 200 runs with almost 95% success rate. And we generally and routinely analyze samples from FFP high molecular weights as well. for leukemias and circulating cell-free DNA. 

We have also had experience working with the Avenio Tumor Tissue and ctDNA Expanded Kits through, again, some projects and collaborations that we had with Roche and other colleagues. And more recently, we've also done projects with the Avenio Tumor Tissue CGP V1. So we were very excited when Roche was releasing the Avenio Tumor Tissue CGP automated kit and be part of this multi-site study. As I said, I'm presenting here on behalf of my colleagues. Our lab is based in Belfast in the UK. We have another lab in Europe, in Leuven in Belgium, who was one of the four labs, as well as Circulogene in the US. Of course, the baseline lab, the benchmark lab, was the Roche Diagnostics Lab in Santa Clara in California. 

Tara has just presented to you the performance of the kit. And of course, the performance of the kit, which is really impressive in over 5,000 samples that we've heard. That was all data that has been generated, sort of, in Roche. And as usual, we tend to sort of give more value to the data that comes from external sites and particularly with multiple sites. So this study that we performed, it's a real-world study in different labs with different settings and different levels of experience. So all of the labs basically run the Avenio-IH system with a tumor tissue CGP automated kit. And in total, we analyzed almost 400 samples.

So this, on the left-hand side, is the distribution of the different cancer types. So there were 48 unique tissue samples across 14 tumor types, so really the most commonly analyzed, tried to represent real practice, lung cancer, breast, colorectal, ovarian, melanoma, et cetera. Basically, in each of the labs, we analyzed 96 FFP DNA samples, so 48 samples in replicate, basically. They were analyzed across four different Avenio-H runs, 24 on each, and they were sequenced in two NovaSeq runs with 48 samples per flow cell. It's a very streamlined process, a very simple process. And the Roche data from the Ford lab was obviously the one used at the baseline to evaluate the call rates. 

So the first thing is there was a very high pass rate across all the different sites, despite some initial sort of teething issues with the edge, which were easily resolved. And there was one issue as well in one of the sequencing runs with... slightly over-cluster, which resulted in lower clusters passing filter. But overall, it was a very high pass rate, just clarifying and showing how easy it is to do a fully automated assay. But more importantly, when we look at the quality control metrics, some of the key metrics that you can see here they were incredibly tight. So if you look from top to bottom and left to right, we basically have the coverage at 100x, which, as you can see in all the four sites, achieved very close to 100%. 

Median coverage was also incredibly tight, slightly lower in site two. That was due to the over clustering issue, but still about three times higher than the minimum required for quality pass. So very, very robust. Equally, the on-target rate was very high, close to 85% in all the four sites. And similarly, the error rates were all below 0.03% and generally sort of around 0.02%. So very, very tight quality metrics, in my opinion, given that this is four independent sites across the world using this assay for the very first time. 

There's another QC metric which I find very interesting and very unique to this assay, which is normalized coverage noise. So, for those of you that are sort of familiar with copy number analysis, you can see here on the top a relatively clean sample with hardly no copy number alterations, probably low tumor content, whilst here you can see a sample also potentially with low tumor content and very low level of copy number alterations, but with significant noise. You can see a very choppy sort of profile. And equally, when in samples with high aneuploidy, like good tumor content with a significant number of alterations, you can see a very clear a clean sample with copy number alterations between a sample that has slightly more noisy data. 

So this assay can actually estimate the normalized sort of coverage noise ratio. And as you can see, in all the four labs, we have basically very close to the one threshold, which is effectively the ideal situation with hardly no noise. So very, very good quality metric to look at potentially, and particularly for copy number alteration analysis. 

Another very, very critical part of this, in my opinion, is the estimation of computational tumor purity. So, as some of you probably know, Foundation Medicine many years ago published their own proprietary algorithm to estimate tumor purity based on sequencing data. And as you can see here, we have different, across the 48 samples that were used, we have different tumor purity from sort of lower of the 20% and up to almost 90%, which again reflects real practice from large tumors with plenty of tumor content, with down to some others which will have less tumor content for different reasons or inflammation. 

And what's very, very important to me is how tight the results are for all the different labs. So, no matter in which parts of the plot you're looking at, all the labs are basically achieving the same value. So that to me, and you can see here the comparison between the Roche data against all the other sites, and it's basically an R-square of 1. So these data basically tells you that the tumor content that you have on your DNA sample is very, very accurate compared to the estimation that we normally have by pathological assessment of the H&E slides. And remember, when we do pathology assessment on the H&E, we're just looking at that in two dimensions. We're not considering the third dimension, which is the depth of the tissue that we're going to be cutting sections to extract DNA, and we're also not considering how accurate the technicians will be at microdissecting. 

So, in this sense, having an internal quality control for the tumor purity inside the DNA that you are using for comprehensive genomic profiling is incredibly useful because it gives you absolute certainty when you do not find alterations, whether that sample is actually genuinely wild type or whether there is a problem with tumor content. 

So, if we just go straight into the results from the multi-site study, the first thing that I want to show you is basically the high positive call rate compared with the data obtained at the Roche lab in California. So as you can see here, all the three other external sites achieved incredibly high positive call rates, sort of close to 99%, basically, goes up to 100% when the variance near the limit of detection were excluded. So you can see here that, for example, the only small blip that you can see, and when I say small, it's very small, it's still close to 98%, is a particular rearrangement, as well as here, you're looking at sort of copy number, and still you're looking at marginal calls, which are obviously calls very close to the limit of detection, which you would always have to report with care and caution. So very, very high call rate of close to 100%. 

And here, we're looking at the reproducibility of variant culling across the different sites. So, running pairwise comparisons of all the different combinations between the Roche site and any of the three external sites and each other, you can see a very, very consistent and robust performance, both for average positive agreement and average negative agreement here, which are basically very close to 100% when you exclude marginal calls near the limit of detection. So again, this shows very, very reproducible and robust data across four different sites using 96 samples on each site.

If we look at other type of alterations, particularly those genomic signatures, which are slightly more complex alterations to detect and where you really require comprehensive genomic profiling tools, we're looking here, again, how reproducible and how robust the data for MSI and TNB, so microsatellite instability and tumor mutation burden. were across all the sites. So obviously, for MSI, which represents a biological feature of the tumor cells, you have basically samples that are either stable or deficient at different levels. And very importantly, again, the data and the results, the score, specific score for each of the four sites for individual samples is all very tight with coefficient of variation below 6%. 

And similar, when we focus on tumor mutation burden, this is more of a continuous variable. Of course, there are samples with 9 mutations per megabase, and 9 1/2, and 10, and 10 1/2. So of course, there'll be cases within the borderline area. But the important thing is, again, all the different sites achieving very consistent, very robust, and very tight scores for TMB with coefficient of variation less than 10%. So, again, very, very robust and reproducible data across the four sites. 

The other type of genomic signatures that are very important and they're becoming more and more relevant lately are HRD signature and genomic loss of heterozygosity. So, again, You can see here a very, very tight distribution of the calls across the different labs. So cases with HSV negative basically are represented here below the 0.7 cutoff. And very clear, all the cases are presented in different labs, analyzing different labs in replicates have very tight results. We have one case here, as you can see, which are borderline, sort of the 0.7. So, of course, again, when you're in borderline marginal calls, you can have different replicates just coming slightly above or below. But for cases that are clearly HIV deficient, there's a significant gap. You can almost fit the palm of your hand in between the negatives and the positives. 

And the same with genomic loss of heterozygosity score, very tight values across the lab for positive cases. and very tight also for negative cases. And in this case, also very clear across the entire cutoff region. So again, very reproducible scores for very complex genomic signatures, which are difficult to detect by conventional tests. 

I think all of that is obviously very good news that using the same samples across four different centers with the same kit, we obtained very high concordance, close to 100%. But I think to me, it's also very important whether those calls that are very tight and very robust are actually correct. So, how can we assess that beyond using the Roche lab in California as the baseline?

 So basically, these are the data comparing the results from all the sites, including Roche as well, against the reference method, which is Foundation Medicine F1CDX. So as you can see here, there's 100% positive percent agreement and almost 100% negative percent agreement. And again, very, very tight across all the labs, across all the different types of alterations. So this is not just single-nucleotide variants. This includes indels, copy number amplifications and loss, rearrangements, fusion genes, and obviously all the other alterations. So very, very important that the results that were obtained across the four sides, which were incredibly robust and reproducible, were also 100% accurate, 99.5% negative predictive value as well. So that's excellent results. 

And if we focus particularly on key actionable biomarkers, which we have in the different tumor types, so EGFR, BRAF, PIK3CA, KRAS, BRCA, et cetera, we can see a 100% positive call rate across all the different types of variants. So not just for single-nucleotide variants and indels, but also for copy number amplification, copy number loss, and fusion gene detection. So very, very robust results for the key actionable biomarkers that we have license indications. If we look also at the more complex alterations, so the gene signatures for MSI, TNB, HRD, and LOH. 

We have also 100% correlation apart from two cases that you can see here. One is tumor mutation burden in a non-small cell lung cancer. And as you can see, this is a case where it basically falls in that cutoff of 10 mutations per megabase. So, of course, sometimes you will have 9.6, sometimes you'll have 10.8, and that's where the 12% comes from. And similarly, in this case here, with HRD signature, is a case that falls bang on the cutoff of 0.7. So again, cases that are difficult to report, no matter what, and that you should be careful anyway. Everything else is 100%. And these are very, very complex genomic signatures that are very difficult to assess with certainty with conventional methods. 

In addition, and finally, we had also data generated across the four sites on four cell lines, two of which represented cell lines with known somatic alterations, in which we basically obtained 100% correlation across all the different sites for all the different mutations and alterations, including copy number gain, copy number loss, as well as fusion gene detection. In both cell lines, but we also analyze reference material, normal reference material, to ensure that was high specificity. So there were two different cell lines. There were no false positive variants detected by any of the four sites. So these are the data generated across four different sites worldwide. 

And the conclusions, as you can see here, is we have now available a simplified and fully walk-away automated system with the Avenio Edge Liquid Handling System. The quality control metrics are incredibly robust. You can have guarantee that wherever you are in the world, you will basically obtain the same level of performance. a high positive call rate of almost 100%. And equally, when we look about the precision across different sites, you have close to 100% average positive agreement and negative agreement across all the different alterations, not just mutations, but copy number and translocation detection as well. 

More importantly, for other different, more complex genomic alterations, we also have a significant correlation with coefficient of variation, less than 14%. And to me, perhaps the more critical aspect is the incredible correlation with accuracy with the reference method, which is, as we've heard, the method that has thousands of publications and the largest amount of FDA approvals for a single genomic profiling assay. And finally, just using commercially available cell lines, we achieve 100% correlation. And I think this is very important for labs out there because this is obviously part of the type of material that you can use in your labs to verify the performance of such an assay in your center. So I'm going to hand over to Sunandini, who's going to wrap up the end of this webinar. Sunandini, please.

Sunandini Chopra, PhD: Thank you, Professor Gonzalez de Castro, for that overview. And thank you to all of you for staying with us for these results. 

So to wrap up, I want to highlight that the Avenio CGP Automated Kit really delivers three key value pillars to our users. So the first one is insight excellence, which is powered by Foundation Medicine's genomic expertise, ensuring high-quality, meaningful results. Second is true walk-away automation, enabled by the Avenio Edge liquid handling system, offering over 24 hours of walk-away time with only 3 and 1/2 hours of total hands-on time. And finally, the third one that is robust performance, which is demonstrated by over 99% overall pass rate and proven compatibility with challenging sample types, including small specimens such as corneal biopsies. So these strengths make the Avenio CGP Automated Kit a powerful solution for labs seeking scalable, reliable, and simplified in-house CGP testing. 

So while today's focus was on the Avenio Tumor Tissue CGP Automated Kit, I do want to highlight that Roche also offers a manual CGP assay as well within the Avenio portfolio. This provides flexibility for our customers and our users to choose the solution that best aligns with their infrastructure, throughput requirements, as well as operational preferences. 

And finally, to conclude, in today's session, we covered some key attributes of the Avenio Tumor Tissue CGP Automated Kit, starting with the streamlined automation with only 3 and 1/2 hours of total hands-on time, enabled by the Avenio Edge Liquid Handling System. Second, proven performance across over 5,000 samples with high sensitivity and specificity. Robust reproducibility, as demonstrated in a global multi-site study, as covered by Professor Gonzalez de Castro. Strong analytical precision and alignment with reference methods. And finally, a solution truly built for real-world CGP, enabling decentralized, simple, and high-quality CGP testing. So together, these trends position the Avenio Tumor Tissue CGP Kits as a trusted and scalable CGP solution for labs worldwide. I do want to thank you for your time. And I'd like to hand it over to Chiara, our moderator.

Kira Curtin: Thank you, Sunandini. As a reminder to our webinar participants, if you have a question, please type it into the Q&A box in the control panel. We'd like to ask attendees also to take a moment after the webinar has ended to take our exit survey and provide us with your feedback. And so now we'll get into the Q&A part of the webinar. And also for this part of the webinar, we're being joined by David Mo, Senior Product Support Specialist at Roche Diagnostics. And the first question that we have is for Sunandini. I think you said that the kit works with NovaSeq. Does it also work with NextSeq 1000 or 2000?

Sunandini Chopra, PhD: Sure. Thanks, Chiara. So the Avenio CGP Automated Kit works with NextSeq 500, 550, as well as the NovaSeq 6000. Regarding NextSeq 1000, 2000, I want to highlight that we are constantly looking at expanding sequencer compatibility, improving our product. So we do have roadmap plans. I recommend folks to be in contact with their local Roche representatives to understand what's in our roadmap and when that will be available.

Kira Curtin: Okay, thank you. And Tara, we have a question for you about one of your slides. I think it's number 16. I don't know why to go back toward that one. The question is, what type of variant was called? And was it a SNP? I just couldn't quite tell.

Tara Love, PhD, CGMBS: So I'm looking at slide 16, and that's about all our data for our studies. But in general, I think for all the information and data that was presented, the variants that were called were in the main four categories of single nucleotide variants, indels, copy number alterations, and rearrangements, along with complex genomic signatures. So, you know, across the board, and maybe David Mo, you can qualify this, but across the board, it's my understanding that these performance results cover all those classes.

David Mo: Yes, that is correct. The defined variant reference set are including soft variants, copy number alteration, rearrangement, and also certain [inaudible]. Thank you.

Kira Curtin: And David, we have another question. I think that's for you. It's about the quality range of the FFPE used in terms of DDCQ or DIN. If 300 nanograms of FFPE is used, what's the minimum quality? Or basically, how bad can FFPE samples be?

David Mo: Yeah, so this assay actually does not use CQ value from qPCR on the DNA integrity number from bioanalyzer or type station to assess the quality of DNA extracted from FFPE. So the input recommendation is 20% or above in terms of tumor nuclei contents, as well as 10 millimeter square or above in terms of tissue surface area.

Kira Curtin: Okay. And for David or Tara, could you go into more detail on the independent reference method? Was it a hybrid capture panel or an amplicon-based method? Sorry.

Tara Love, PhD, CGMBS: I can answer that. It was a hybrid capture method, and David, you know, referred to it at the end of his talk, where he said there are thousands of publications and FDA approvals, you know, it's very robust, so. That's the answer. Hybrid capture.

Kira Curtin: Hybrid capture.

Tara Love, PhD, CGMBS: Yeah.

Kira Curtin: Oh, let's see. And Sunandini, is it 8-plex or can one sample be processed at a time?

Sunandini Chopra, PhD: Yeah, so with this CGP automated kit, users can do 12 or 24 samples at one time in the Avenio Edge system. However, for sequencing, they can sequence all the way from up to 12 samples in a NextSeq, 500, 550, and 24 to 48 in a NOVA-Seq 6,000. But on the sequencer side, it really depends on the lab's appetite. They can sequence as few as they like. Obviously, the cost per sample would increase. But as far as the liquid handling system is concerned, we do 12 to 24 at a time. Users can choose fewer samples as well. It will just increase the price per sample for analysis.

Kira Curtin: Okay. And David, for the multi-site testing results that you presented, what was the input amount of FFPE that you used?

David Gonzalez de Castro, PhD: So we used, there was a range of input from FFPE, so varying from the 40 nanograms to 300 nanograms, so covering the range that I think we've discussed before, where the assay has shown performance. So the minimum recommended, I guess, is 40 nanograms.

Kira Curtin: Okay. Thank you. And let's see, Tara, did your analysis of genomic alterations include large rearrangements and homozygous deletions in HRR genes? I can't hear you if you are talking. I'm sorry.

Tara Love, PhD, CGMBS: You mean you can't hear me when I'm on mute? No. Yeah. So, you know, I'm going to ask David Mo to help me with this one. You know, the HRD-SIG biomarker that we have really does take into account a ton of different large rearrangements and copy number alterations and things in the genome. And so to that, you know, I can say to the extent for that, that covers sort of the question that's being asked. But as far as large rearrangements and homozygous deletions in particular, you know, in the HRR genes outside of our HRD SIG score, I'm just wanting David Mo to kind of qualify or kind of help me out with this one.

David Mo: Nope.

Kira Curtin: Yes, yeah.

David Mo: So I mean, for HRD signature, we actually use some machine learning algorithms to train on basically what we define as HRD true positive versus HRD true negative, and then for HRD true positive, like we do use the [inaudible] mutation as defined as our true positive. So, in some sense, it does include the mutation in the HRR genes, if that answered the question.

Kira Curtin: I think so. Yeah. And for Sunandini, are the bioinformatics included, and what is different versus other kits in the market?

Sunandini Chopra, PhD: Absolutely. So the bioinformatics truly is one of the key USPs of our CGP assay. So the bioinformatics is included as part of the Avenio Tumor Tissue CGP kits, both the manual and the automated. And it is powered by Foundation Medicine. So we offer our users the Foundation One analysis platform. So, of course, the question about the difference is compared to other assays. I mean, we leverage the Foundation One analysis platform, which I mentioned in my talk is based on over 10 years of really strong leadership in the CGP space. And we have a team of scientists constantly working to improve and update that bioinformatics pipeline. So we are offering, really, the brain power of foundation medicine through our bioinformatics to our users. Okay.

Kira Curtin: Uh, and, David, what, uh, tertiary analysis are you using with this assay?

David Gonzalez de Castro, PhD: So, um, we, um, have access to Navify mutation profiler, which I think is the perfect combination for this assay. So, um, obviously it's part of the Roche portfolio. And I think it just interfaced very, very nicely. So you can basically take the output from the bioinformatics pipeline strain to navify mutation profiler, and this automatically will produce a report with all the actionable alterations and selection for clinical trials immediately. So it just requires no substantial additional analysis.

Kira Curtin: OK. And let's see. For Sunandini, if we're a manual library preparation, can one sample be done?

Sunandini Chopra, PhD: Yeah, so using the Avenio Tumor Tissue CGP Kit V2, which is a manual workflow, users can go down to one to two samples as well. I don't know if David Mo wants to add anything, because we have customers who've done that when they are trying the assay initially.

David Mo: Yeah, one sample is fine.

Kira Curtin: OK. And also for you, Sunandini, how many reactions can be done with a manual CGP kit?

Sunandini Chopra, PhD: Yeah, so a manual CGP assay includes 24 reactions in one kit. So users can do up to 24 in the library preparation. However, when it comes to sequencing, we recommend up to 12 samples in a high throughput flow cell in the NEXseq 500-550.

Kira Curtin: Okay, and also for you, can it be run on an S4 NovaSeq 6000 flow cell?

Sunandini Chopra, PhD: Yeah, so Roche internal team has not tested the S4 flow cells. So we have tested the SP and the S1 flow cell. So yeah, I can only speak to what our team has done rigorous testing on.

Kira Curtin: Okay. And another one for you, Sunandini. Are there automated solutions for the Avenio Tumor Tissue Expanded Kit?

Sunandini Chopra, PhD: So, actually, unfortunately, at the moment, we do not have an automated edge TDF for the Avenio Expanded Kits. However, it's great feedback. And we are constantly updating our roadmap for our assays and for our automation. So yeah, please be in touch with the local Roche representative for any more information on upcoming features and roadmap. OK.

Kira Curtin: Let's see. This one is for David. Since you've used the Avenio CGP Automated Kit as well as the prior Avenio CGP Kit assay, can you please share your experience and recommendations?

David Gonzalez de Castro, PhD: So I have to say they're both very impressive. And obviously, we have a team that, you know, thankfully, they're very experienced in the lab, so they've had a lot of experience with hybrid capture assays for many years. And obviously, when we introduced the Avenio CGP V1 manual, it was fairly easy for them because they have that much experience. 

But I have to say it's a very, very onerous process compared to running fully automated on the edge. I mean, it literally is, you load a plate of DNA samples or the lysate if you want to do the extraction on deck. You press play. And the following day, you take the final pool and put it in the sequencer. It literally is that even I could run it, possibly on a good day. So the main feedback is that the results are very similar in terms of the chemistry, of course, and analysis, but the performance is so much more robust as we have seen in the data today, because you completely eliminate all the variables to just let the machine do its job very well.

Kira Curtin: That looks like we're coming up on all the time that we have for today. So I'd like to take the time now to thank our speakers, David Gonzalez De Castro, Sunandini Chopra, Tara Love, and David Mo, as well as our sponsor, Roche. If we didn't have time to get to your question, we'll try to follow up with our experts. And as a reminder, please look out for the survey after you log out to provide us with your feedback. Also, if you missed any part of this webinar or you want to listen to it again, an archived version will be emailed to all attendees. Thank you for joining us for this Genome Webinar.