Hi everyone, welcome to this session of Perkin-Helmer Genomics. I hope everyone is staying safe and healthy and thank you for joining this session. My name is Madhuri Hegde and I am going to be talking about whole genome sequencing. Are we ready to define this as our new clinical standard? At Perkin-Helmer Genomics, we have been doing whole genome sequencing on a clinical basis for the past three years alongside gene panels, whole exome sequencing, and some of the other traditional methodologies such as targeted sequencing as well. We have gathered a significant amount of experience doing whole genome sequencing on a clinical basis. So, I hope I can convince everyone today that we are really in that era now where whole genome sequencing is close to becoming a standard for clinical care. Before we jump into Perkin-Helmer Genomics setup, I want to spend a few minutes on looking at where we, how far we have come with next generation sequencing. This slide is a quick summary of the timeline of sequencing, of how we went from using southern blots, Sanger sequencing, to this new era of next generation sequencing, massively parallel sequencing, and all the different technologies and equipment that is available for us to use in clinical care now. This is still going to continue to evolve as new methods get developed. But this is really where we are today, and it's quite an exciting time to be doing genomic testing. In Perkin-Helmer Genomics, we essentially took this concept further. And what we have set up is everything around the box. And what I mean by that is that we use a sequencer from Illumina, which is a NOVA-66000. We have 10 sequencers in our laboratory. And that box is surrounded by Perkin-Helmer equipment, starting from DNA extraction using the QMAGIC-360, a lab chip nucleic acid analyzer for sample QC, automated library prep using the robotics of Janus G3, and then, again, a lab chip GX for library QC, followed by next generation sequencing, and then an in-house developed software analysis program, which we call ODIN, but which is surrounded, again, by a full suite of different application programs. This helps us control the entire pipeline within the laboratory and allows for efficient and easy workflow to generate whole genome sequencing. The automation tools are entirely within the Perkin-Helmer family, which really gives us that advantage of controlling the end-to-end process of sample preparation. So, one of the questions which sort of will be good to sort of set right now here is that how many samples can we process on a NOVA-seq or whole genomes? So, on the S2 flow cell, we can do 48 copy number genomes, which we do at a 5x coverage, which essentially replaces microarrays, and 1140x whole genomes. And then, on the S4, 144 copy number genomes and 2240x whole genomes using the S4. The one important thing here is that we are not doing 30x genomes in our lab, we are doing 40x. And there is a reason for that, because it allows for optimal variant detection, including single nucleotide variation and inclusion. As I mentioned, we have also put together an entire ecosystem, if I can call it, around which the informatics revolves, which includes ODIN, which really is assembling the data, annotating the variants, and doing variant interpretation. But before and after that, to complete the entire suite is the LIMS system, which manages the workflow, the batching of the samples, also the reporting system from which ODIN generates a PDF quality report and sends it back to the LIMS, along with quality assurance, version control, and internal audits. This is entirely written and developed at Perkin Elmer Genomics, and what you can see at the bottom is tools which come external, which is the HGMD OMM databases, which are commercially licensed, and then we also use the Biodiscovery software for copy number variation detection and the fDNA tool for face recognition and, at times, variant prioritization. So let's start with looking at the age-old question of, is whole genome sequencing really beyond whole exome sequencing? We obviously know that the first difference, which is the exome is about 1 to 2 percent of the genome, but the key here is in the method used for genome sequencing versus exome sequencing, where the genome sequencing is a truly unbiased sequencing, a sort of PCR-free, which is called a tagmented or tagmentation approach. And we are covering the entire genome, including the full gene sequence of the exonic or the coding sequence and the intronic sequence, whereas in exome or the gene panels, we are targeting plus-minus 50 bases. So if a deep intronic variant is shown to be pathogenic by RNA sequencing, it will not be detected by exome sequencing. Nowadays, the bioinformatics tools have evolved enough that we can actually call copy number variation from exome sequencing, but genome sequencing really gives you that edge in that aspect as well, and I'm going to show you an example where exome did not pick up a copy number variation. It was so easy to detect it with genome sequencing. And then comes this next three rows here, which you can see is the ability to detect repeats, differentiate pseudogenes, uniparental disomy, and then also calling regions of homozygosity, which makes genome more comprehensive. We have not gone beyond that right now into talking into the methylation aspect of it, and I'm sure that is going to come soon in clinical testing. The other way to look at this is can we advance the exome towards the genome to make it more cost effective, because there are some still areas or regions of the world or also within U.S. where whole genome sequencing may not be as cost effective, depending on how the payment is going or the reimbursement is working. But the idea really is that we are capturing beyond the exome in when we are doing whole genome sequencing, and some of the data, published data, has shown that it is possible to capture approximately 15 percent disease-causing variants that are in the non-coding regions of the genome itself. But the idea that you're capturing the entire intronic sequence is critical here, especially in recessive conditions, when you have one pathogenic variant, and at times in dominant conditions too, where the clinical diagnosis is there, but you have not seen a variant in the coding. The other important aspect of this entire process of doing PCR-free genomes is the uniformity of the data. So what you see on this slide here is the comparison between a 40X genome and a 100X exome, and you can see this highs and lows in the exome, whereas the genome data at 40X shows unbiased, more or less uniform sequencing. The next question for anyone doing clinical testing is, how does this compare with the ACMG59 gene list, the genes in the SGMD database, RefSeq database? And what you can see here is side-by-side comparison of genes and exomes in these databases, and whole genome sequencing does afford a higher coverage compared to exome sequencing. Now if we take this data and convert it into number of variants, a little bit more reflective of what we see as lab directors or even data analysts, and what you are attention to on this slide is the number of variants seen in ACMG, SGMD, RefSeq, and the total number of variants in exome and genome sequencing. And you can see that the exonic variants are slightly higher, that is largely because the number of dropouts from a genome are significantly lower, dropout regions within an exome, but also the ability to call the variants and detect the variants is higher. And then if you compare the total number of variants, you expect to see more variants in genome sequencing just because you are sequencing the entire gene. So this is a direct head-to-head comparison of exome to genome sequencing. Let's look at number of variants you can detect using different sample. On this slide, what you see is a comparison of variant calls from dried blood spot card, extracted DNA, and put through genome sequencing. Genomic DNA could be extracted using variety of different methods, saliva, and whole blood. And you can see that the data is very comparable. Purkinje-Elmer is really known for a very long time for newborn screening. And today, newborn screening is done in about 155 countries on about 50 million babies, which is pretty phenomenal. In the U.S., 4 million babies are born every year approximately, and almost all babies go through the newborn screening approach where a dried blood spot card is collected and tested for the, at a minimum, the rust panel of 29 genes. Dried blood spot cards can be transported easily, and at Purkinje-Elmer Genomics, because we have global labs in India, China, Malaysia, Sweden, we get samples from all over the world coming to us, and dried blood spot card becomes a very easy mode of transportation. The number of variants, the average number of variants is very indicative of the data quality we are producing. But also, if you reflect on the average of percent basis at a Q30, the data is of high quality, and we can use it to interpret for clinical conditions. Another important thing which we have all struggled with over the years is the coverage of GC-rich regions. And what you can see on this slide on the left-hand side is where the exome sequencing drops off in the coverage of the GC regions, whereas genome sequencing is still continuing to give you that coverage as the percent GC increases. On the right-hand side, you see the obvious gap there at a 100x exome versus a 40x genome where you're filling that gap, and in this case, we are showing you the example from KLF2 gene exome 2, which has a 77% GC in exome 2. If you look at exome level detection, copy number detection, again, genome sequencing does have an edge over exome sequencing just because of the fact that you have uniform coverage across the region, and the baiting is not applicable to genome sequencing because you're sequencing basically the naive DNA, whereas in exome sequencing, you're using baits to capture that region to interpret at a clinical level. Some of the additional things of advantages of doing genome sequencing is the confidence in calling short tandem repeats. Here what I'm showing you on this slide is an example of the PABPN1 gene, and what you can see is that with whole genome sequencing, the edge, as I mentioned, is when you compare the data, you can very easily resolve a false positive using genome sequencing, and you can actually be very confident in those calls. If you take the same sort of table and look at it into a more visual aspect of it, what you can see on this slide is this gap here in exome sequencing compared to genome sequencing with the uniformity of the sequencing across these regions. So this past year, we spent a significant time on validating a number of repeat disorders, and earlier this year, we launched detection of short tandem repeat disorders using whole genome sequencing. The list is shown here on this slide. The focus has been that we can rule out when we detect two alleles or heterozygosity. At times when the repeats are apparently homozygous, we cannot rule in which requires that if the clinical phenotype is close to any of these genes, then additional testing or confirmatory testing will be required, such as a triple prime PCR or doing southern blot in certain cases when the repeats are extremely high. But this allows you as a starting point at an entry point to repeat disorders, which is critical for many physicians so that they are getting this comprehensive view and then they can decide what to do next. Other extended advantage of genome sequencing is SMA. And what you can see on this slide here is the algorithm which differentiates between SMN1, SMN2 using the copy number ratios of distinct SNPs and our side-by-side comparison and confirmation using an alternate technology, such as real-time PCR, MLPA, and we have also used commercial kits to do the confirmatory testing between repeats of copy numbers detected by genome sequencing to what we got through confirmation, and you can see that it matches. On the left-hand side, you can see that we can very clearly detect the positive, the carrier, and the normal range. This has come through extensive internal validation in our lab so that we have been able to establish this algorithm to make the call. The example shown here is of these two variants, which are absolutely critical to making this call to differentiate between SMN1 and SMN2. These are the intronic variants and a promoter variant, which you can very easily detect and call from genome sequencing. So let's look at some data now of how we have been able to use genome sequencing in a clinical setting. So the obvious question that comes to mind is, what is the diagnosis? This slide shows you is the differences between microarray, exome, and genome sequencing, and you can see our pickup has been significantly higher. We have held back on publishing our data so far, but we are in the process now of bringing our data and submitting it, because we have done a significant amount of cases, and also including the different sample types. Because a lot of data has already been published on whole genome sequencing, it was really important for us to show these advances in our population. If we look at the first 400-odd cases, what you can see is the diagnostic yield on one side, but the type of the variants we have detected using genome sequencing in the pediatric population, shown in purple, and then the adult population, shown in green. These single nucleotide variation is obviously the highest, followed by the copy number variation, but the difference in the pediatric and the adult population is quite obvious on this slide. The number of WSIS, or the variance of uncertain significance, is very interesting here. It largely depends on the clinical presentation, the details given to the laboratory, and then thereby our interpretation, and whether a singleton was done, or trio. And I'll show you the data of differences between singleton and trios as well. But this is taking that data further, and deep diving into the pediatric cohort versus the adult cohort, and you can see that our diagnostic yield in the pediatric cohort is higher, whereas a variant of uncertain significance is pretty much similar. This really largely depends on the need for functional assays, and need for additional testing on families. But if we look at side-by-side, the singleton versus the trio, the trio does offer a very slight advantage, but the difference here is at the bottom, where you see the number of singleton cases versus the trio cases we have done in our lab. And the problem here being that there are many cases which are probably likely pathogenic or pathogenic, but trios were not done. And the trio situation will significantly alter if that follow-up testing is done. And we can actually move that needle on that variant itself. Now showing some data on experiences with healthy genomes, this gives you the comparison of the pediatric and the adult cohort. We do healthy genome sequencing largely through our cod blood-cod tissue channel, which is a wire-cod channel. So we have done newborns and their parents. And what you can see here is a type of the categories which we have sort of detected pathogenic variants in, which ranges from cancer syndromes, neurodegeneration, endocrinology, skeletal dysplasia, to write down to more of the sort of deafness and the skin disorders that are documented within OMIM. The type of diagnostic findings really range from aneuploidy, microdeletions, duplications, intragenic variants, uniparental disomy, and then unsuspected unbalanced translocations. There's a significant amount of carrier detection as well that happens, absence of heterozygosity, and then the variants of unknown significance, which are not reported, largely because they need further work. And we do reclassification, internally reclassification, and if a variant is reclassified, we will go back and issue a new clinical plan. So let's quickly look at some examples of the application of genome sequencing in a clinical setting and all the sort of claims I have made in the last 20 minutes, if they do they actually translate into clinical testing or not. The first case here is a five-year-old male with a suspected diagnosis of MPS2, or Hunter syndrome. This particular case was biochemically diagnosed and confirmed to be MPS2. They did Sanger sequencing, MLPA, and which both were negative, and the physician wanted to follow up and see if we can, if we do genome sequencing, if we can detect the variant. Now, the first question that would come to mind is why not sequence just the MPS2 gene completely? That would have required us to make a new library, a new assay, to just sequence the MPS2 or the IDS gene, rather, whereas genome sequencing, we were able to mask the rest of the data and only look at the IDS gene data, and far more cheaper than actually just starting to develop a new assay which sequences only the IDS gene. So this is what we found. We found an intronic variant in the IDS gene. You can see it is in the intervening sequence 6, exon 6, and it is minus 72 base pairs upstream. This variant was originally classified as a variant of uncertain significance and now moved into the likely pathogenic criteria. This family has been, has gone through RNA sequencing, and this has been confirmed to be pathogenic. The second example is a previously missed diagnosis. Now this is a very interesting example because if you look at the family history, it's quite complex and comprehensive here in, starting from neurodegeneration, dysmorphic features to multiple other features noted on the slide, but if you look at the testing history, the biochemical testing, Fragile X, karyotyping, FISH, single gene sequencing for some obvious genes, mitochondrial genome sequencing, and exome sequencing, which was done in another laboratory and only one variant was detected, which really fit the phenotype. Now exome sequencing, most clinical labs today are able to do copy number variation, but this particular exome sequencing done elsewhere did not detect the copy number variation or rather it's a missed example of a missed variant by exome sequencing. You see this on the slide here that we were able to detect deletions of exons 8 and 9 using genome sequencing, which was previously missed by exome sequencing. So again, exome sequencing can detect copy number variation, but genome sequencing does have that edge due to the ability to do unbiased sequencing. What you see here is just a little bit of a more zoomed in view of the maternally inherited variant in exon 2 and the paternally inherited variant, which is a deletion variant of exon 8 and 9 in the NRS2 gene. This solved the sort of the mystery for this particular case and they were able to get a diagnosis, which matched the clinical presentation in this individual. The next example is enhanced copy number detection, which is also a very interesting example. This is a 6-year-old male with history of intellectual disability, speech problem, abnormal hands and legs, and there was no reported family history or consanguinity. What you notice here is a very interesting aspect of the advantages you can have with genome sequencing. So it looks like a homozygous deletion with a larger heterozygous deletion, which suggests that both copies of the D5M gene are involved here. So just to go over, what we see is a deletion of exons 6 to 13, and then a single exon deletion of exon 13 only. This makes it really interesting and difficult to resolve when you are doing the data analysis. We have previously published another such example of a deletion hidden under a duplication for Crabbe disease, which is even more challenging because of the duplication you see a neutralization of copy number variation. So in this case, as I mentioned, we have a deletion within a deletion on two separate alleles, deletion of exon 6 to 13, and then a single exon deletion of exon 13 only. The last example I'm going to show you is cytogenomics, where can you actually replace the microarray? This is a two-year-old female referred for developmental delay and a range of other features. And as we know that now microarray is considered the first-year testing for constitutional disorders, we had last year published a paper of replacing the microarray with a 5X genome, which we call the copy number genome, which is not only cheaper, but more comprehensive and has a better delineation of the breakpoint. So what you see here is a karyotype analysis, and you can see this bit here on 7q36. The results of copy number genome sequencing are shown here, where we detected the pathogenic copy number variant, which is a gain, and all the genes listed on the clinical report itself. But if you look at it from a more of a microarray-based view, it is very clear that you can see that low-pass genome sequencing is a very good alternative for microarrays. Now genome sequencing is going to capture everything, right, from the large microdeletions duplications to single exon level. But if a microdeletion duplication type situation is suspected, then you can replace the microarray with copy number genome. This shows you a more zoomed-in figure of the 21 megabase interstitial copy number gain of 7q31.32q34. Very easy to see, and this encompasses about 105 omim genes, which were there on the previous slides. Obviously, I did not expect you to read that gene list, but many of these genes are known to be causative of a genetic disorder. So I hope I have shown you all the power of genome sequencing, how we are moving in that direction of genome sequencing, what we can do to enhance panels and exomes to move towards genome sequencing, understanding the drawbacks of exome sequencing, how we can enhance the exome, but then we can take the genome forward and stretch and extend our ability to do bioinformatic analysis using genome sequencing. I will stop, and I'll be happy to take any questions. Thank you.

whole genome sequencing

clinical standard

Perkin-Helmer Genomics

exome sequencing

copy number variations

short tandem repeats

microarray analysis