Skip to main content

The European Human Genetics Conference 2025

View the recording of our talk at the European Human Genetics Conference from May 24-27, 2025 in Milan, Italy.

Great, excellent. Thank you so much. All right, let's get started. So let's start then with a quick introduction to our company. So Blueprint Genetics is a Quest Diagnostics company and we deliver high quality genetic testing to the global clinical community. We have over 250 employees and we serve over 4000 clinicians in 70 countries and we have core operations in Helsinki in Finland. We offer a range of NGS based test types depending on the customer's needs. But let's focus for now on whole exome sequencing. So whole exome sequencing aims to sequence the protein coding regions of all known genes. And it's important to note that this is distinct from clinical exome where you sequence disease associated genes only with a Wes you aim to sequence all known genes regardless of whether they have a current gene disease association. So Wes covers approximately 1 to 2% of our our DNA because obviously the vast majority of our genome is non coding. But nevertheless Wes is a very powerful diagnostic tool because there is no restriction on what genes you sequence. And of course at the moment the vast majority of known disease causing variants are in the coding regions or the splice regions. So, so the regions that Wes does cover whole XM sequencing is very good for patients with complex phenotypes and an unclear clinical diagnosis. So patients like this, it may have been difficult to know what gene panel to choose for that patient, what targeted tests to do. So Wes is potentially very good for patients like this. We also find that Wes is very useful for patients who have had a previous inconclusive genetic testing result. And we find that's typically because there have been some technical limitations to the previous testing the patient may have had, may have an atypical presentation. So again, it may have been difficult to know what targeted testing to do for that patient. And also often we see on Wes if there has been previously inconclusive testing that the molecular diagnosis ends up being in a gene with a very recent gene disease association and genes like that may not be on a panel. So these are the three elements that I want to focus on today in talking about the elements in general of an optimized whole exome sequencing assay. So there are many elements of an optimized whole exome sequencing assay. And today I will focus first of all on the technical side on the inclusion of clinically relevant non coding variants on where's optimized CMV detection. Then I'll talk about the importance of making sure that the gene disease associations that you have in the system are are current and up to date and the importance of skilled interpretation that's done in such a way that gives us the best chance of detecting relevant variants in patients even if their phenotype is atypical. First of all though, let's start by talking about trios and why we recommend that when you run a whole exome sequencing test that you do it as a trio where possible for the best possible diagnostic yield. So a trio Wes is kind of the typical Wes setup where you have affected pro band and to unaffected parents if that is the family setup that you have. And we really recommend running the Wes this way. And the reason for that is that it really facilitates the interpretation. It enables de Novo variant detection. This is really important because in our experience approximately half of of molecular diagnosis on where cases are de Novo variants. It's also important for determining the phase of variants. If you have two variants in the same gene gives you valuable recurrence risk information and a more accurate variant classification straight away without necessarily needing any further family member testing. So, so it can give you all the testing that you that you need in one step potentially. So let's take a look at the case to illustrate the value of doing Wes analysis as a trio. And by the way, all the cases that I show you today are based on real cases from the Blueprint Genetics Archive, but with some details changed or removed to protect patient confidentiality. So this case involved a small child with psychomotor delay, hypertonia and mild dysmorphic features and there were no affected family members and the analysis was run as a Wes trio, so the affected pro band and to unaffected parents. This patient actually had several rare heterozygous mis sense variants in genes associated with autosomal dominant neurodevelopmental disorders that showed overlap with the patient's reported phenotype. And this is not at all an unusual situation in a WESA analysis that you would see several rare variants like this that that that could fit with what's described for the patient. And this illustrates the difficulty sometimes with a pro band only analysis that if you don't have the inheritance information, it can make the interpretation quite challenging, especially when you have somewhat nonspecific and limited clinical information like we have with this case. So in this sort of situation, what you may end up having to do is report multiple variants of uncertain significance. So you get a longer report and then more family member testing to do to try to resolve those variants of uncertain significance. However, this this case was luckily a trio and this makes life a lot easier. Because we had the parental data, we could straight away see that the KMT 2A2A variant was de Novo, so not detected in either parent. And the other missense variants were inherited from unaffected heterozygous parents, allowing us to rule those out. So the KMT 2A variant was the only variant reported for this patient and according to our classification scheme and the de Novo occurrence, it was classified as likely pathogenic. So this patient received a molecular diagnosis of KMT 2A related disease due to a de Novo variant which comes with important information for genetic counselling and and recurrence risk. So doing the analysis here as a trio really made the difference. It facilitated the interpretation, led to a simpler report, gave you the immediate likely pathogenic classification without needing to do any further family member testing. So the TRIO approach can be very powerful and all the cases that I will show you today actually were run as a TRIO Wes. OK, so let's move on now to some technical aspects of an optimized Wes, starting with the value of adding clinically relevant non coding variants to to the Wes assay. So let's take a moment to think about where in a gene we find disease causing variants. The majority of known disease causing variants are in the coding regions of genes. So of course the exons and also at the splice regions, these are the regions that Wes typically covers. But of course we know these days that there are also disease causing variants in non coding regions such as deep within introns and other non coding regions such as UTRS and and promoters. For this reason we target specific non coding variants on our whole exome sequencing assay. We do this by seeding with additional capture oligos to target specific non coding regions where there are potentially clinically relevant variants. So that means variants that have been reported in patients in the literature or are very nearby to variants that have been reported. So, so these variants can can be interpreted. That's really important. This method enables us to detect almost 2000 potentially clinically relevant non coding variants on our Wes assay. And this is really valuable because based on our internal data about one in 20 Wes molecular diagnosis involve a non coding variant. So it's really valuable to be able to capture these on whole exome sequencing. Let's take a look at a case where this was this was valuable. This case involved a child with muscular dystrophy and again the analysis was run as a trio. OES analysis detected a known pathogenic variant in the Col. 6A1 gene. This was seen to be de Novo because we had the parental data and this variant had been reported in multiple patients with autosomal dominant collagen 6 related congenital muscular dystrophy. So this was a direct match with the patient's reported phenotype. This variant is located deep within intron 11 of Col. 6A1 and kind of a standard out-of-the-box where's assay would typically cover the exons and going maybe 20 base pairs into the intron to pick up the splice region variants. So it wouldn't detect this deep intronic variant, but we can specifically target this deep intronic region where this variant is located, allowing us to detect it on whole exome sequencing. So in this case, the patient then received the molecular diagnosis of Col. 6A1 related disease due to a de Novo variant. And what really made the difference here was the inclusion of clinically relevant non coding variants on the assay. Let's move on now to take a look at the importance of optimised copy number variant detection. And also at the same time, I'll talk about the value of having skilled interpretation incorporated into your Wes analysis. So copy number variant detection from NGS data is done at Blueprint Genetics using a method method called read depth mapping, which is where you compare the number of reads you observe in a particular region with the number of sequencing reads you would expect to see in that region. And then a significant difference between the two generates ACMV call. And we use three complementary bioinformatic tools to call copy number variants from the data. First of all, a tool called CMV kit which calls larger deletions and duplications and then a small an in house algorithm which is especially designed to detect small deletions because the other tools are not not so great at doing that. And these small one or two Exxon deletions are are clinically very important. And then we also use a breakpoint detection script which calls breakpoints in the NGS data to potentially alert you to the presence of a of a CNV there. We find that deep and uniform coverage is really key to getting good CNV detection from from from the Wes and it's well published in the literature that copy number variants are are really relevant clinically and our in house data backs that up. We see that one in seven diagnostic W cases involve a CMV and about 10% of those are CMVS under 1000 base pairs. So, so especially these very small CMVS are really important to be able to detect those. So let's take a look at the case. This case involved a deceased fetus with multiple fetal anomalies including short shortening of long bones and micro Nathia and a skeletal dysplasia panel previously had been negative. The Wes analysis detected compound heterozygous variants in the COG 5 gene, a known pathogenic sequence variant, and a single exon deletion. Bioelic variants in COG 5 have been reported to cause a extremely rare disorder termed congenital disorder of glycosylation type 2 I, and This disorder has an extremely varied clinical spectrum and and and and and the spectrum is still evolving. Bioelic variants in this gene were originally reported in a small number of adult patients with mild developmental delay and ataxia, but then the spectrum soon evolved to include intellectual disability, short stature, microcephaly and dysmorphic features in further patients with proven COG 5 deficiency. And then more recently, retinal degeneration and skeletal dysplasia were added to to the spectrum. And then very shortly before this case was interpreted, the first severe fetal phenotype was reported in a fetus homozygous for a known pathogenic variant in in COG 5. And the phenotype of this published fetus was very similar to what was described in in our fetal case. So how about the classification of of the variants? So I said the sequence variant was a known pathogenic variant and then we had the single exon deletion. This was actually a deletion of an in frame exon encoding only 37 amino acids. So because it's very small and it's in frame, we can't classify it as a predicted loss of function variant. We can't apply that that classification point. There was at least one pathogenic MIS sense variant in the Exxon involved reported in the literature suggesting that that this affected Exxon is functionally important. But nevertheless this is a very tricky classification this small in frame deletion. So in the classification we took into account the phenotype match and the fact that there was a very good match between our fetal case and the fetus reported in the literature and also with some of the elements of what was described in in the disorder post. Natally as well, the deletion was absent in in in reference populations and affected and assumed functionally relevant exon. If we think about the reported missense variance in that exon. And then the really key piece of evidence was that we can see, we could see the deletion was in trans with a pathogenic sequence variant in the same gene in the affected foetus. So this is not easy, but based on this evidence, we classified the deletion as likely pathogenic. So this family then received a molecular diagnosis of COG 5 related disease, which importantly comes with a 25% recurrence risk for any further offspring of the of these parents. And there were a few things that made the difference in this case. So first of all, optimize CMV analysis to enable the detection of this very small deletion, but also careful and skilled variant interpretation to be able to accurately classify this small in frame deletion. And also to recognize the importance of these, the relevance of these variants in this patient whose whose phenotype is not, is not typical this fetal phenotype, which is only reported very recently. So it's important to have skilled interpretation to take all the literature into account and especially the most recent recent literature. OK, so moving on now to the final, final elements of an optimised where's that? I'll talk about today the importance of having up to date gene disease associations in the system. So a gene disease association refers to the relationship between a gene and disease. And when you're interpreting and classifying variants, it's really important also to take into account the strength of the evidence supporting the relationship with disease for that that particular gene. And at Blueprint Genetics, we classify gene disease associations using a framework which is based on the Clingen framework. And in our Wes analysis, we aim to focus mostly on genes with a moderate or above gene disease association because anything below that is not clinically relevant at that time and there isn't enough evidence at that time to support that gene disease association. And of course, we want to concentrate on clinical clinically relevant findings in what we report out in our whole exome sequencing reports. So of course that means though that having up to date and correct gene disease associations in in whatever your system you're using is really crucial. So let's take a look at a case where this was really important. This case involved a young adult who had died suddenly and was found to have dilated cardiomyopathy. The exome analysis identified compound heterozygous variants in the NRAP gene, a frameshift variant and a missense variant. However, there is no gene disease association for the NRAP gene on the OMIM OMIM database on Orphanet, not on Gentifen, and there is no available curations for NRAP either on Clingen or by Gen. CC. However, in 2021, bialelic loss of function in NRAP was reported as a cause of recessive dilated cardiomyopathy. There were 11 unrelated pro bands reported who had DCM with bialelic or two NRAP variants and there was significant enrichment of bialelic NRAP variants seen in patients with DCM. And there's also a CRISPR knockout zebra fish that has a cardiac phenotype. Based on this evidence, we consider that the gene disease association between NRAP and recessive DCM is strong. So looking at the variants identified in this case, both of them according to our classification scheme can be classified as pathogenic. They have both been reported in multiple patients and meet all all the the points required for for pathogenic classification. So this family then received a molecular diagnosis of NRAP related recessive DCM for this unfortunately deceased patient, which comes with the very important information that there is a 25% risk to any of their their siblings. And what made the difference here is having up to date gene disease associations in the system and to make sure the gene disease associations that are in whatever tools you're using are based on reputable databases, but also the primary literature. Because as we see here, there are some, some gene disease associations missing on databases. And, and this is particularly important these days, I think because looking through the, the program for this conference this weekend, there is a lot of talks and information about AI driven variant prioritization and automated methods. And those, those tools are only as good as the the gene disease associations that are that are programmed into them. And in fact, we, we tested a couple of whole genome analysis tools using AI variant interpretation. We tested those in house and both of those missed this NRAP diagnosis because the gene disease association for NRAP was not, not in the system. So really important to make sure that the correct gene disease associations are included in whatever tools you're using. OK, So those are the elements of an optimized where's assay trios where possible, technical optimization and the inclusion of current gene disease associations and also skilled interpretation. And there's one element that I haven't talked about yet, which is inclusion of the mitochondrial genome. And and that's because it deserves a talk all of its own. So I will hand over now to my colleague Raquel, who's going to share our experience with including the mitochondrial genome on on whole exome sequencing. And thanks to you all for listening. Thank you. Thank you so much, Kirsty. That was really, really insightful. And and I said we'll take the questions in the end. So please already put them on paper so you remember that. Before we continue, I just want to say that there's a few places still in the front row. So if you are standing, please feel free to come here and you can sit down as well to eat. As already mentioned. Next we will look what happens when you go beyond the nucleus, so adding the mitochondrial analysis to the to the XO man. And for that, I'm super happy to present our next speaker, Doctor Raquel Pereskora, who is also a geneticist in the clinical interpretation team specialized in, in whole exome sequencing. She has a PhD in retinal dystrophy and and has a 13 year experience in, in, in genetic Human Genetics. So please welcome. Thank you, Salvin, and good afternoon, everyone, and thanks for joining us today. I'm going to talk about mitochondrial diseases and the importance of integrating mitochondrial DNA analysis into a whole exome sequencing approach. Mitochondrial diseases are very complex, both clinically and genetically, and that's why including mitochondrial DNA testing into the workflow can make a real difference in identifying the underlying cause. So first a brief introduction to mitochondrial diseases. They are a group of genetic disorders caused by mitochondrial dysfunction and therefore defects in oxidative phosphorylation which lead to impaired energy production. And one of the key features is that they can affect different systems or organs, especially the high energy demanding ones such as the brain, muscles or heart. And another hallmark is that is the significant phenotypic variability and the progressive course of the symptoms complicating their diagnosis. So these conditions can present with a wide spectrum of symptoms and some of the common ones include ataxia, optic neuropathy, hero loss, cardiomyopathy, muscle weakness or diabetes. This disease can arise from variants in mitochondrial DNA and these are maternally inherited as mitochondria are passed down exclusively from the mother. And they can arise also from variants in nuclear DNA, in nuclear genes that are affecting mitochondrial elements or empty DNA maintenance. And these conditions can follow an autosomal recessive, autosomal dominant or X linked inheritance patterns. So the mitochondrial genome has distinctive characteristics that can directly impact how we interpret mitochondrial variants. The mitochondrial DNA is a small circular double stranded DNA molecule. There are 37 genes including subunits of respiratory chains, complexes, and also rrnas and trnas that are essential for mitochondrial protein synthesis. The mitochondrial genome lacks Internet regions and about 93% of its sequence is coding. Except for this sorry non coding region, the mitochondrial genetic code differs slightly from the nuclear code, which is important to consider when annotating or interpreting variants. Another important aspect is that it's polyploid, meaning that its cell contains multiple mitochondria and its mitochondria contains multiple copies of mitochondrial DNA. So this leads to the concepts of homoplasmy where all of the mass majority of the copies are identical or heteroplasmy where there is a mixture of normal and mutated Mt DNA in a cell. And the heteroplasma level can vary not only between individuals but also between different tissues in the same individual. And this is known as a tissue specificity. And this can significantly affect the disease expression. So now let's talk about some technical insights. Currently, the next generation sequencing is also the new gold standard for Mt DNA sequencing and but what are the main challenges we face in mitochondrial DNA testing? So in the one hand, we have the technical challenges that include ensuring a uniform coverage and high sensitivity. Alignment and mapping processes can be also complex and may lead to false positives or misinterpretation of variants due to the nuclear mitochondrial DNA segments, which are mitochondrial fragments that have been integrated into the nuclear genome. And also it's important reliable detection of low etoplasmia variants. And on the other hand, we have, we face also some interpretation during some challenges, sorry, during the interpretation process. So there is a very limited population data available for mitochondrial variants. The literature can be inaccurate. Many publications lack some critical information, such as the heteroplasma levels, the tissues tested, or the detailed clinical descriptions. Also the viability in phenotypes or heteroplasma levels that can be different even among individuals covering the same variants make the interpretation challenging. And also we also need some specific classification guidelines that are fully adapted to mitochondrial variants. So at Blueprint our whole some sequencing include a high quality mitochondrial DNA analysis. The mitochondrial genome is sequenced in parallel in with nuclear genome. The whole empty DNA genome is covered with uniform coverage. All types of variance SMBs, Intels and CMBS are detected and we have a specific bioinformatic pipeline to detect CMBS. Our essay has a high sensitivity and good with depth to reliably detect the low etterplasmic variance which are which could be clinically relevant. So these are our performance statistics and as you can see we have membrid depth of more than 9000 reads. And regarding heteroplasma detection capabilities, we are able to detect SMBs in theirs or CMBS with a high sensitivity at a very low level heteroplasma. So also a blueprint we have, we have a large in house population data which support more accurate variant interpretation. We carry out extensive literature to review and we we carefully interpret not only variants, but also the teachers tested the phenotypes, the heteroplasma levels or disease thresholds and we also have our systematic classification scheme for mitochondrial variants. So why it's important to include mitochondrial DNA analysis in West. So as Kirsty mentioned, many of the cases that are referred for exams present with complex phenotypes or unclear clinical diagnosis or a typical presentations that don't clearly point to specific syndrome or a specific disease. And these are the precisely the kind of cases where mitochondrial diseases might be hidden or misdiagnosed. As we have mentioned, the mitochondrial diseases are challenging as they have high phenotypic variability. There is a progressive course of the symptoms and and they can mimic other conditions and making the diagnostic process particularly complex. So now I want to show you some internal data of mitochondrial findings in West. So mitochondrial findings contributed to 1.3 to 2% of diagnostic cases. In 70% of the cases, pathogenic or lilipathogenic mitochondrial variant was the only diagnostic finding. In 20% of the cases, the mitochondrial variant was found together with nuclear variant, and both were lightly contributing to the phenotype. And in 10% of the cases, the mitochondrial variant's role was uncertain. And on the right hand side, you can see the most prevalent teens and we are still gathering some W data to accurately evaluate the diagnostic impact. But the addition of mitochondrial genome to our panels resulted in a 1.1 increase in diagnostic deal. So these are the four recurrent disease causing variants found in our core as well as the heteroplasma levels found and the clinical phenotypes reported in our patients. And one interesting observation was the wide range of heteroplasma levels found associated with the M3243A2G variant and the variability in the phenotypes reported in our patients. And this variability highlights the complexity in the interpretation process. So what does this mean for for patients? Now I'd like to share with you 3 interesting cases that highlight the importance of the potential impact of including mitochondrial DNA analysis in West, but also the importance of the interpretation and reporting process. So the first case is a baby who presented with epilepsy, apnea and suspected bradycardia, and the patient had a previous epilepsy panel performed at another laboratory, which was inconclusive. And the patient was referred for waste triad blueprint together with the unaffected parents. So there were no diagnostic variants in nuclear genes, but when we analyzed the mitochondrial genome, we found this variant in MTAT P6. The heteroplasma was around 55% and as you can see in the picture the variant was absent in mothers sample which could be consistent with the novocurrence. The variant was absent in population databases and it creates A premature subcodon removing 23% of the protein. And interestingly, the variant was previously reported in the literature and it was seen in a patient with LAY syndrome. And the variant occur at 21% heteroplasmy in blood sample. And the variant was also detected by other laboratories and submitted in Klimar. So what about the phenotype? Variants in MTA TP6 are associated with highly variable phenotypes including late syndrome, neuropathy, ataxia, readiness, pimentosa, upper motor neurone disease, as well as other isolated manifestations. And the most frequent symptoms include developmental delay, epilepsy, seizures, ataxia, neuropathy as well as respiratory abnormalities including AMNIA and some less typical symptoms such as cardiomyopathy and conduction defects or lactic acidosis. And as you can see highlighted in red, our patients features could be explained by this variant. And what about classification? So in summary, our variant was absent in databases. It creates A premature subcodon. It's been reported in a patient with mitochondrial disease. There is an establishing disease association and there's a good match with our patients manifestations. The heteroplasma level found was consistent with what's being reported in the literature and also the variant was likely the noble so we could classify the variant as pathogenic. So this patient received diagnosis of MTA TP6 mitochondrial disease and the variant was absent in mother's sample, which is consistent with the Novo occurrence. And this result enabled to provide an appropriate genetic counseling and management of the patient and family. And what made the difference in this in this case first, inclusion of mitochondrial DNA in West which is especially relevant in cases like this very young patient in whom mitochondrial disease would wouldn't necessarily have been expected. Our high sensitivity and rhythm assay allow us also to indicate that this variant is slightly been Oval and also having our systematic classification scheme and doing a careful interpretation.  And the second case is a young patient with hearing and vision loss, diabetes, brain MRI abnormalities and mild cognition concerns. And there was no family history reported. This patient had multiple previous genetic tests, including a mitochondrial test, which and all of them were uninformative. So when we analyzed this case, we didn't find any diagnostic burden in the nuclear gene, but what we found in the mitochondrial genome. So as you can see in the picture, we were able to identify 2 heteroplasmy mitochondrial large divisions, one at 10% heteroplasmy and the other one at 30% heteroplasmy. So we also were able to detect breakpoints in both of them. And 1 includes 24 genes and the other 113 genes, none of them were previously reported in the literature, but several similar overlapping deletions were reported in patients in the literature. So we were able to classify both as pathogenic. So large mitochondrial deletions are associated with mitochondrial DNA deletion syndromes which includes overlapping clinical phenotypes which are characterized by retinopathy, cardiac conduction defects, ataxia, cognitive decline, sensorineural heal loss, endocrinopathies or brain MRI abnormalities. And our patients manifestations are highlighted as well in red here and you can see that there is a really good clinical correlation. So we were able to finally confirmed mitochondrial DNA deletion in this patient after a lifelong diagnostic odyssey. The large mitochondrial indegations typically occurred in Ovo and this result allowed to provide an appropriate genetic counseling to the patient and what made the difference in in this case. So first, the detection of mitochondrial DN, A/C, MB, S in Exxon and this is thanks to our customized clinical bioinformatics analysis and also having a high sensitivity. I read the assay allow us to detect the large mitochondrial with deletion with low heteroplasma. And the last case is a baby who presented with microcephaly, global developmental delay, hypertonia, some brain abnormalities and small optic nerves. This patient also had some previous and informative genetic tests and the family history include a knowledge sibling with small head side but normal brain examination. So what we found in the nuclear gene were these two heterozygous missions variants in ASNS. Gene 1 was classified as pathogenic and the other one as slightly pathogenic, and the phase was of the variant was unknown at that time. Variants in this gene are associated with aspiring synthesis deficiency, which is characterized by congenital microcephaly, severe developmental delay, hypotonia, ******* quadriplegia and some brain MRI findings. So as you can see, most of our patients features could be explained by by this disease except for the small optic nerves. So we continue our analysis to find an additional variant that could explain that feature. And what we found is this homoplasmic pathogenic variant in MTND 6, which is a well established variant in levers hereditary optic neuropathy or lung. And this disease is associated is characterized by visual filler, reduced visual acuity, centroscoptoma, retinal vascular issues, optic atrophy, optic dysfunction as well as other extraocular features. So the optic nerve issues in our patient could be explained by this variant. But we wonder why our patient doesn't present the typical lawn manifestations. So we investigate it further the novarian in the literature. And what we found is that this variant is associated with best long term visual outcome and it causes less severe phenotype with visual recovery seen in some patients. And the patients typically present an onset that range from adolescents to adulthood. And also there is low penetrance reported in the literature and some specific approach may influence the disease pressure. So we got molecular findings in both a nuclear and mitochondrial genome. So we could have a possible double diagnosis here for our patient. The mitochondrial finding also support an early ophthalmologic follow up which is significantly relevant in this patient as it's a baby and lung symptoms may appear later in life and these results enable to provide an appropriate genetic counseling to the patient. So what made the difference in this case? So mainly a careful interpretation by doing a comprehensive, by an assessment in the literature, checking phenotypes reported, the tissues tested, heteroplasmes or disease thresholds or even penetrans, and also by elucidating genotype, phenotype correlations. And of course, of all of this information was reflected in a high quality report. So that's all from a part. A huge thanks to the whole BPG team for the effort, dedication and passion. And thank you for listening. Thank you so much, Raquel. That was really interesting. Next we will move towards Whole Genome and the EXO, what you just heard about from Kirsten and Raquel, it's probably still a more more comprehensive test than the genomes we can find on the market, but we all know it's time to move towards that. And next, I'm super proud to welcome to the stage, our medical doctor and Co founder of Blueprint Genetics doctor, medical doctor Yuva Koskevor to kind of a bit reveal our thoughts and approaches to watch how to analyze the whole Gen. We all have the data, but the question for everyone is like, what do we do with the whole genome data? How do we interpret that? Then what's the best way to go around it? So please. All right, good afternoon, ladies and gentlemen from my side as well. My name is Yuva Koskevor and my background medicine and research and and I'm also Co founder of Blueprint Genetics. Today my topic is about clinical genome and how we analyse the data with prioritisation tools. And I also tell you something about history in this section. So going back about 20 to 30 years, so we started the filtering and prioritisation steps in those days. So first steps, first tests for like chenotyping a single scene testing and typically we see 12:50 to 10 variance and we didn't have good population databases neither we had like systematic classification approaches and therefore we many times classify the variance as this is causing in relation to phenotype. But fortunately we have developed a lot from those days. In Step 2, we use typically osteis that contain several hundreds of genes and then we slice out the data from that using filtering by gene Orion consequence and population frequencies. And in the early days we had really limited population data from thousand 1000 genomes and and a small database that has improved by time. In step three, we used mostly clinical EXOM and EXOM backbones to slice out larger panels and EXOM interpretations. In that step, we used different filtering approaches already improving from Step 2. We also used mutation databases to develop better design for the assays to detect like let's say deep internal variants also, so that variants that are difficult to find, so we put them to the assay and we have also already made interpretation in the system for those. Then in Step 4, we do genome level analysis using exomes and genomes. It's already very challenging. Typically we see something between 35,000 var ions to 4.5 million var ions and it sets a very high demand for next steps. What comes to var ion QC and var ion interpretation, If we calculate how much time it would take to analyse 1 genome so that you would put one second for each variant, it would be 156 days doing 8 hours work. So everybody can understand it's too costly, so we need to filter the data. We can't anymore filter it only with the population frequencies and variant consequence. We need to be more clever. Also we need to remember that cost of testing is is reduced and the complexity increase during the last decades. So we need to do something clever and one answer to make it better is MIMIIAI related prioritisation tools. So with these slides I want to highlight four things. So of course we have a need for larger test because we still live without molecular diagnosis with many cases. So what clinical genome cannot bring us, it improves detection of structural abriance, but it's also set a new requirement. So there is a possibility to call very large number of them. So we need to have good step forward on QC and also they are difficult to filter with population frequency because those databases are still at infancy. So many times you need to have your own databases to enable good filtering with genome. It also provide possibility for reanalysis and you need good tools for that so that you don't keep on analysing the same things again and again, but you are able to see what has changed after the first analysis. Then going the big issue is the number of the variants. So it's sets the biggest cost of the whole genome at the moment and it's not only interpretation but it's variants you see. So if we have millions of variants and we need to put lots of time to decide which one are true positives and which are false positives, it's difficult. And if you think about what Illumina has done in this section, they have developed rock and pipeline and they want it to be streamlined that enables massive amounts of interpretations and not so that we stuck with one genome and tracking pipeline is handling a bit different way those variants which are in the low mapping region. So if you think about the patients with hearing loss due to stereosciling gene stereosciling mutations, we have a drop in that region if you use tracking pipeline. So but why that's that's a feature is that we can't ripples of millions of Orion's we need to have very high predictive value that they are true positives. So going to the today's topic, it's genome level Orion prioritisation. So what you can do there, you can internally develop tools for that. If you have a high volume centre or then you're going to use third party tools. There are many, many third party tools in the market, at least 20. Typically these companies are relatively small having like between 10 to 30 employees and they have good know how in software side a bit limited understanding many times for clinical interpretation side and senatics also some LAX in quality requirements of these kind of things because it may have like a regulatory things that prevent using certain assets in So the reads and CDPR may be a problem in EU if your data is leaving EU and so on. So one of the most promising tools there is a illuminous imaging is good fabric by fabric synomic that was acquired by C&DX and Franklin. Franklin Synox is good both by Kiagan and EVI and Synom just to list few of them. But I said many other tools are also good. But I need to say that all of these have limitations and nobody can claim that they can perfectly analyse 10 at the moment. If you want to see how these tools perform, you need to take a look how they actually fit to your clinical interpretation purposes. You need to understand the bio uniform techs and software engineering architecture that they use and you need to understand how they can be updated and how financially situation the company is so that can you trust your future for that kind of vendor. Then you need to understand the quality and regulatory sides because you need to do in Europe. Europe, you need to have a software as a medical device that is very soon and there may be also security risk in these softwares but you need to investigate all of these if you want to use them. So we have come through many of these vendors and starting from the spaces what we need to take a look. So better you can do the variant QC in the system or do you need to done it before you put those variants in the system or can you still manage the variant call through a false side in in the tool? Then you need to know what kind of para and types they are supported. Is it SNVSCNV, structural variants, mitochondria variants, repeats and whether they can combine variant combination from different para and categories. Then you need to understand how the HPO workflow is functioning and what kind of further creations they have done, because probably all of you know that have done HBO things. There are lots of mistakes in there. Then we need to understand what kind of ML or AI prioritisation tool they are using in the system, how it performs. Then we need to see how we can work with the variance after the prioritisation, what's the reanalysis function, how you can integrate the tool to your existing systems and mutilated databases and so on. And we have done evaluation of certain vendors with our cases. Of course, we select very difficult cases for those because if it would be straightforward, all of these tools would pick if we just put let's say nonsense variety for myosin, Piney protein, C3, these tools still find it, it's not a problem. But now I show you some difficult cases and by these detection performance you it doesn't reflect these tools performance. So what we did, we insected these variants for the 1000 Senos space and data there's 12 SNVS, 1 structural variant and two CNVS. Besides of these we insected also other variants that have high battles. This is discovering with our all of our system that is internal tool for variant. I would say this is predicting, predicting and then we put them using track and pipeline. There then be be Bams and VCFS were uploaded to the platform and then analyzed with certain HBO terms and and patients sex and so on. And this didn't include anything any information for our knowledge base or CD accuration so and so on. So if he would do full, full integration, then the yield would be better. Also with this data, I don't want to try to put any of the vendor in the bad light. So I disclose that maybe none of the labs that I saw the name by name are not part of this evaluation. So it's still all the all the systems have limitations. So I have now the case is case one is from my OSAP, it's homocycles nonsense Varian and it was among top one, top five. None of the vendors and it was 30% of the vendors put it in the top ten. And one of the problem of, of that gene is that it's still quite recently discovered synthesis association also it it's within the complex transcription unit. So this varion may be called in three different genes and it may be said difficulties to find varion pairs. Case 2 is in draft varion. It's something similar that Kirsty saw today. So it was actually missed by all vendors because of of synthesis association miss. And then case 3 is Calmodulin one missense variant of course like CPVT, Daslang PT syndrome and it was found 70% of the cases in top five variants and top ten in all. So very well done. Case 4 is double diagnosis so PTCH 1 neurodevelopment disorder and known as syndrome pair and it was found 70% in top five and top 10 variants. Then next case is desmiccline 2 variant Peres homocycles Dystein typically associated with dominant disease and very few variants are found to associate with recessive phenotypes and and it was also found by 70% of the cases in top five and top 10 paravaritic size variants. Then case 6 is telomeropathy variant. It's a heterocycles one finished founder for pulmonary fibrosis. But this variant was very low quality in the data and it was missed missed by most laboratories probably due to low variant quality. Then case 7 is ARMRP homozygous variant. The non coding site it's known by the chinic variant causing like cardiolis hair I will place here and it was found 70% in both categories. Case 8 is RBM 8/2 what ion pair missense variant and interonic variant. This is causing like thrombocytopenia and absent radio syndrome and it was found also 70% of the cases. And the problem is with the low penetrance of the interonic variants. It doesn't cause disease somocycos but pair with the severe variant it cause severe phenotype. Case 9 is UBA 1 missense variant that was missed by most laboratories and I have on the slide of that. And then case 10 is ABCA 4 that relates to star card disease, retinal dystrophy disease and it's a pair of nonsense variant and mist sense and it was missed by all cases because the mist sense is very common and it's it's hypermorphic variant. And many times these kind of AI prioritisation tools have problems with, with hypermorphic variants. So a bit more this is the end up variant. So it was missed the TDA. So it, it's not surprised that it was missed. And also the missing side is, is has very high splice area score. So it's, it's relevant in that sense as well. So then it's the telomeropathic variant. So it was missed due to low quality and the UPA one is causing Vexer syndrome that is immunological and hematological phenotype. And many times these tools and CDA inheritance modes are a bit problematic when the phenotype is mostly actually linked recessive. But in many cases actually it cause phenotype in certain parts, certain portion of females as well. And in this case 4% of Vexer syndrome patients are females and it was not noticed by most, most vendors and this missed APCA 4 missions variant is, is has very many classification in green bar and it's very, very common. This is Kosi variant actually. Then going to our internal efforts for variant patozenicity scoring or prediction. We have developed Oliver during last five years and it tries to separate between patrotenic and P9 variants and it's fits for SNVS and Intels. Simply what Oliver does it it generates score for its variants. Something between zero to 1 and 0 means that it's likely more B, 9 and one that it's. It's Patrotenic and we have tried vast number of different parameters when we develop all of our. And final model contest more than 50 different 50 different parameters using the common mass like in silica prediction conservation, all frequencies, loss of function data, splicing prediction, variant consequences, length of Intel variant, position in protein, clean water classification and review status and inheritors models etcetera. With the data. When we developed this tool, we tried different machine learning models and I don't reveal here here today, but we end up the use finally. But different models have different yields. So this is something we have submitted it's not published yet, but this is comparing all over with the existing variance sets. These are big sets of between 3000 to 20,000 variants from Cream's data, Cream data sets and we compare Oliver with other, other tools like Clean Thread, DOJ, FAT, ATM and Revel. And four out of five cases, however, is better than other prioritisation tools in predicting pathogenicity. Whenever, whenever we do this kind of things, it's the most important for novel variants. And this is comparing these insidical tools with for novel variants. And you can see that all of our performs very well comparing to let's say clean pred, CAD, reverend other tools. So it's, it seems that we have developed good, good tool. But despite of that, it still misses some sometimes variants. So in here we retrospectively take a look how it performs in certain set of cases and we found that 1.3% of cases it misses the variant. So we need, we can't have that high number of missing variants. So we develop additional features to rescue those variants that are missed by the machine learning tool. So you can use of course your previous classifications that you don't enable those variants to be filtered out. Then you can of course combine it with phenotype. If you approach the stenotype first in prioritisation, you can enforce certain variant pairs, you can have a certain scene specific thresholds when the model doesn't fit fit well to the certain scene. And of course when when we go further, there may be need to validate for new variant types and and and train it better. Especially the internal variants are very important in the future when we start to use scenes in larger, larger scale. But it's a show that all of all of the Orient prioritization tools, they have problems. You need to understand the risk when you incorporate to to your system and you can rescue and rescue certain misses that be done by doing additional checks. That was all from my side. Thank you. OK, thank you so much. Johan, you have will also be tomorrow around 12:00, I think at the booth. So, so please welcome there to ask and discuss more. Our last but not least presentation presentator is Lottakoskevoa. Lottakoskina, sorry, who will kind of lead us into screening testing. We have a couple of minutes after, but but hope you have the time to still listening to that. Thank you, Solvei and good afternoon to everyone from my behalf as well. My name is Lottakoskinen. And for the last topic of this session, I'm going to discuss genetic proactive screening tests of healthy individuals so we can actually use the experience and knowledge that we gain from analysing a patient cases. Also to help unaffected currently healthy individuals to learn about the risk of becoming sick of a genetic disorder. And typically, only medically actionable conditions are included in this kind of testing to make sure that there are disease management or treatment options available when needed. So there are many types of genetics screening tests available, but our focus is on disorders that are primarily caused by pathogenic variants in single genes. Proactive genetic screening tests are intended for healthy adult individuals who don't have a significant family history of a genetic condition and WHO want to know about the risk of certain genetic conditions and take actions, according to the findings. Another reason for pursuing proactive genetic testing would be a concern of not knowing disease history in the family so well. So in proactive genetics screening, of course gene selection is very critical and in our offering and in our test design, we have followed professional practice recommendations that are available. So US Centers for Disease Control and Prevention and also ACMG have published lists of genetic disorders that would where early detection and intervention are associated with improved outcomes. So in addition to following this practice professional practice recommendations, we have done an additional review of literature for medically actionable adult onset disorders with well established gene disease association, evidence of moderate high disease risk and availability of well established disease risk management and or treatment options. So then in addition to careful gene selection, also reliable sequencing technology is very important also in screening tests. And inclusion of CNVS in addition to SNVS can bring additional value because in many of the genes that are being tested, there are recurrent pathogenic deletions and applications. And also careful variant correlation is very important in genetic testing. As we know and based on our experience when analyzing global populations, we know that new pathogenic variants are being still found in this also in these genes that are very well established disease genes already. And one important aspects in the interpretation of these genes is that many of them are associated with more than one disorder.  And there can be also different inheritance modes and disease mechanisms. And the interpretation should cover all these different aspects, while the reporting should focus on the medically actionable findings. So based on the data by us and by others, 10 to 15% of tested individuals have clinically actionable findings. And in our data the most common finding have been has been the factor 5 Leiden variant that is the recurrent pathogenic variant that's associated with autosomal dominant increased risk of venous thromboembolism. And the most common gene with reportable findings was the check to gene which is associated with moderate increase in cancer risk. So with based on these kind of findings, then a person may elect a different lifestyle or for example look for a more recurrent health check UPS according to the findings. So then I have an example of a finding that from a positive from a comprehensive cardiology screen test at US. So in this case there was heterozygous pathogenic variant in the KC and Q1 gene. And this variant is a well known pathogenic variant that's associated with autosomal dominant long QT syndrome and it's particularly common among the Finnish patients with this syndrome. The phenotype associated the is wide widely variable and is associated with with the incompletely prenatal so that about 30% of heterocyclic carriers actually have clinical manifestations, but those manifestations can be lethal. So they are actually established recommendations for individuals who carry pathogenic variants in this gene. So beta blockers are indicated for these persons and also they, they are recommended to avoid drugs that we don't predispose to duty prolongation and also heavy exercise is not recommended. And with also with proactive testing and with positive findings, it's important to consider that the findings can have implications also to close family members who are also at high risk of being carriers of these variants and may also want to know about their disease risk and get tested. And it's important to remember that genetic counselling is always recommended after genetic testing, also in the in the screening context. So being mindful of the time, I will skip the last example I had as a positive finding in proactive screening and jump to the summary of this presentation. So, proactive genetic screening tests can help to identify risk for chronic diseases and can support in taking steps towards preventing, delaying or alleviating chronic illness. Careful gene selection, reliable sequencing technology combining SNV and CNV detection and also careful and up to date interpretation to provide medically actionable results. The key elements in high quality proactive screening tests. And with that, I'd like to acknowledge our passionate and dedicated theme at Blueprints and ethics. Thank you thank you so much lot. And maybe rounding up where we started that transparency and quality have always been at the core of what Blueprint does and and we also wanted to bring it to screening, which is a bit of another kind of approach, but we feel like that's really can be a life changing test. So why isn't it equally important there? So, so maybe ending with that note and then any questions now there is good time for those who can say the microphone is there. So, so please, please from any of the topics. Hi, just one question about the mitochondrial presentation, Raquel, right. What about the tissues? Which tissue do you consider as most relevant, for example for mitochondrial variance? Yeah, well, a blueprint molecular testing is done usually in blood sample, saliva or bucal swab samples. But we also perform the the analysis in DNA extracted for different tissues. So in mitochondrial genomes, as you know, probably muscle is the best option to test. So we accept DNA extracted from muscles or other different issues, skin and urine, Yes. So yes, what about the case, I think #2 with the variant not found in the model, would you recommend then also testing other tissues? Yeah. That's why I said that it's likely adrenal ovarian and also that it's absent in this mother sample like in this tissue, because I think we can safely say that it's real adrenal ovarian without checking other tissues. I think it's important to really confirm the adrenal occurrence. Thank you. Very nice talks. Thank you. Thank you. Great questions. Question about the screening test, The screen tests are direct to patient features or there is a, so our screening tests are offered for healthcare providers. OK. And another question, do you consider to do like a opportunistic screen test for the trio exome sequencing when it's a child involved? I don't know the opportunistic screening test for parents in the trio based exome sequencing, Do you mean secondary finding reporting? No, I mean you, you do the trio based exam for a child and then you do the screen test for autosomic recessive diseases for the parents apart from the report. Do you consider that or or not? That's not part of our usual offering. We provide then reproductive screening tests. I mean like opportunistic like a additional feature for the for the family. Go ahead. Yeah. So for for parents, we would provide the ACMC secondary finding reporting, but then analysis of additional recessive carriership of additional recessive conditions would not be part of our whole exome process. We provide additionally these reproductive screening tests if couples are interested in in the risk of being carriers of the same recessive disorders. Thank you. Thank you. Very good question. That as well. Any, any further questions? All right, the panelists will still stay here. So if you want to come afterwards to to kind of discuss them, that's of course as well. And we will be at the booth booth for the rest of the day. So thank you so much for our great presenters, really fascinating talks and thank you for everyone staying, staying to the end. Thank you.