Welcome everyone to the ACMG Genomics Case Conference. The series has been planned by the ACMG Education Committee. Each month, the conference will rotate among laboratories that perform genome scale sequencing and the clinicians involved in the cases. Use of all case materials that will be presented have been formally consented for use with the parent patient or the guardians. I'm Jill Rosenfeld-Milkery. I'm coming to you today from Baylor College of Medicine, and we're speaking on behalf of the Undiagnosed Diseases Network clinical site here at Baylor. So this activity is intended for educational purposes only. Aside from myself speaking today, we have Dr. David Murdoch, Dr. Carlos Bacino, and Dr. Chuan Chow, all from here at Baylor working with the UDN. Here are our financial disclosures. Dr. Bacino has research support from BioMarin and Ascendis, royalties from UpToDate, and is a consultant for Best Doctors. I have salary support from Baylor Genetics Laboratories. Dr. Murdoch and Dr. Chow have nothing to disclose. These are the planning committee members' disclosures. This session is approved for one CME credit. And in order to claim credit for this activity, you will be asked to complete an evaluation within the ACMG Genetics Academy. And you'll complete this at the end of the, complete the evaluation form at the end of the session. These are the overall learning objectives for the case conference series. And then specifically today, we have some learning objectives centered around RNA sequencing. In order to ask a question, we're going to be taking questions live at the end of the seminar. Type in your questions in the Q&A box, so not the chat box, there's a separate Q&A box. And you also have the ability to upvote questions. So if there's a question that you really want answered, there's a little thumbs up icon that you can click on to upvote a question. So now I'll turn things over to Dr. Murdoch to start our content. Thank you. Thank you, Jill. Yes, this is David Murdoch. And so the first point I wanted to make is, I think this is a really good question. how do we think about RNA sequencing within diagnostics in general? So it really comes back to the central dogma of biology, where we have our DNA that goes to RNA, which goes to protein. So at the DNA level, we are talking about studies like exome sequencing, genome sequencing, chromosomal microarray, gene panels, et cetera. Whereas at the RNA level, we're talking about RNA sequencing, so sometimes called transcriptome sequencing. And in this, we're mainly looking for variation in gene expression and gene splicing that can inform and help us to interpret variants that we detect at the DNA level. And then protein level studies are things like metabolomics and other functional studies that can also be important and helpful for us to interpret variants that we find at the DNA level. So what can RNA-Seq specifically tell us? Well, it can tell us a number of different things. It can tell us what are the consequences of intronic variants, which is something very important, especially now with genome sequencing that we're doing. For example, if there's an effect on splicing. It can tell us about nonsense-mediated decay, whether it's occurring or whether it is not occurring. And that's mainly through looking or detecting any decreased expression in particular genes. It can tell us about skewed X inactivation, which can be very helpful in interpreting variants. And then also it can help inform and perform a pathway and downstream analyses that would not be possible with without RNA sequencing. So how is RNA-Seq performed? Well, it's performed in a way that's somewhat similar to how we do DNA sequencing with some important distinctions. So one of them is your tissue source. So this is very critical with RNA sequencing because as you all are probably well aware, different genes are expressed at different levels in different tissues. For example, brain, muscle, fibroblast, blood, et cetera. And so it's important that your final decision for a tissue to perform RNA sequencing on is representative of your genes of interest or your phenotypes of interest. So once you do that, once you do choose and isolate your tissue, you can then extract your RNA from that. And it's processed, it's purified, and a number of different steps are taken. And then it's put on the sequencer. Usually it's subjected to standard next generation sequencing. Then the reads are generated, and then they're aligned to a reference sequence, much like for exome or genome. But this is different in that you use a transcriptome-aware aligner. So where it differs in large part is the subsequent analysis. So how you actually analyze your RNA-seq information. So traditionally, a traditional approach to RNA-seq analysis relies on the identification of candidate variants, typically from an exome or genome, which is followed by the time-consuming manual review of the transcriptome for any effect on splicing or expression. And ultimately, this may lead to the identification of a causative variant. And we'll give you an example of this technique, this candidate variant approach later on. So in contrast, an RNA-seq-directed approach detects outliers in splicing or expression across the entire cohort. And this information can then be combined with an individual's exome or genome data to quickly identify the causative variant and make a diagnosis. The strategy also serves to identify candidate novel disease genes for additional evaluation. And while the candidate variant approach that I spoke of before is effective, there are a number of limitations. First, as we transition from exome to genome, the sheer number of candidate variants grows tremendously, which increases the time required to mainly curate every possible effect in the transcriptome. Second, the selection of many candidate splicing variants is dependent on bioinformatic predictions, which still have variable performance and accuracy. Third, that approach requires the identification in the initial genomic analysis. And lastly, since it's that approach, the candidate variant approach prioritizes known disease genes, it's less suited for novel gene discovery. And so using an RNA-seq-directed approach can overcome some of those limitations. So to answer the question about what kind of tissue is better for clinically relevant genes, we analyzed expression profiles of genes in 16 different disease classes, reflecting the varied phenotypes of patients evaluated at the Baylor UDN site, where the primary phenotype is actually neurologic. And in 15 of 16 cases or classes, fibroblast-derived RNA had significantly higher number of well-expressed genes compared to whole blood, including genes that are associated with neurodevelopmental phenotypes that are shown here in red. And we also did a principal component analysis that demonstrated notably better consistency in fibroblast seen here in red than in whole blood here in blue. And although two distinct clusters are present, the fibroblast data shows much less variability, in part due to the variety of cells that are present in a blood sample. And this suggested that fibroblast RNA-seq is preferable to whole blood for detecting differences in gene expression that have a biological basis and may be clinically relevant. And in fact, these findings prompted our decision to prioritize skin biopsies on probands for RNA-seq. And with that, I will hand this over for a description of our first case. Thank you, David. Thanks, everyone. What I would like to do is show you some, start with one interesting case in the clinic. See if I can get to it. This is actually an interesting patient because it presented with multiple different issues and there were multiple abnormal variables. So presented at three years of age. He had a number of congenital anomalies including hypospadias, ventricular septal defect, the PDA, a butterfly vertebra and a sensorineural hearing loss. The pregnancy, there were a number of issues going on in the pregnancy. The PSD was diagnosed prenatally and also there was pericardial effusion and a UGR. NIPS was done and was non-informative. But also there were a lot of different issues like exposure, the mom was vaping and she actually was exposed to Paxil and other pills due to anxiety, including gabapentin. The baby was born and also had poor apgarus one, four and six at one, five and 10 minutes. And it was small at birth, which already created some concerns and noted to have dysmorphic features, have problems feeding, and in fact necessitated a GE button. Also lots of issues about the family history, this maternal half-brother that has a child diagnosed with bacterial. There is a father-sister with left lip and there were other issues including a maternal half-sibling that died at 36 weeks intrauterine with fetal anomalies and also was small and had cardiac defect and possible gastroschisis. There was a mother's and with Down syndrome and bipolar disorder and depression in the family. As you can see now, all of these combination of prenatal, perinatal and family history were kind of concerning. The medical history is a global developmental delay for this child. And it was rolling over by 12 months and by age three years, he wasn't even sitting independently. He was able to babble and barely use some picture communication exchange. I told you that the fittings were by a G-tube and had severe constipation and ended up receiving cochlear implants. So this is a challenging question. This is one of my clinic visits and you can see that his weight and head circumference were all below the normal. He has microbaric encephalitis, deep set eyes, sparse eyebrows, mid-face hypoplasia, a broad, very peculiar nose. The ears are borderline low set and small with thick helices. Has a small mouth and a high palate but no palatal abnormalities. Otherwise, long philtrum, broad neck, and a low set nipple. And you can see that he has a G pattern. He's very slender and has this sort of interesting built appearance. The exam shows small phallus and hypostadias. So the differential was very broad from chromosomal abnormalities to perhaps some single gene disorders or X-linked disorders. One of the things that I consider at the time was, I recall, 1P minus and even coughing ceres but nothing really panned out. In the hands, he has fifth finger clenodactyly, small nails, sort of this postulated appearance of the tips. And he has a mild syndactyly between the third and the fourth and minimal second to third. And this particular patchy appearance in the skin. A number of studies were done. He has synoptic pelvic left kidney, the heart anomalies, the vertebral anomalies, as I show you, normal array, normal panel for sterols. And obviously, I'm not including a lot of metabolic evaluations that were normal. Provan exome and trio exome was essentially non-informative except for a couple of recessive disorders. And I believe David goes next. Yes, that's right. So for this case, the RNA-seq directed approach revealed decreased expression and abnormal splicing in the PQBP1 gene in the proband in whole blood. So PQBP1 is interesting because it's a gene on the X chromosome that's associated with Renn-Penning syndrome. And if we look at the actual alignment, so this is an IGV, something called a sashimi plot where the proband is in red and then two controls are in blue and green. We can clearly see the creation of a pseudo exome between the normal exomes three and four in the proband. And as well as we see some other splicing abnormalities, some intron retention more distally. And what's interesting is that this is an out of frame deletion that would be predicted to cause an early frame shift and lead to nonsense mediated decay, which is what we were detecting by the decreased expression from the RNA-seq analysis. And so we did have a clinical genome that had been ordered through the UDN that was negative, was unremarkable. But we went back and we did a reanalysis given the RNA-seq findings. And we found that there was a deep intronic variant in the PQBP1 gene that was hemizygous in the proband. He's a male. And again, this is on the X chromosome that was inherited from his heterozygous mother. And what's notable is that this deep intronic variant activates a cryptic splice donor site, which was leading to the pseudo-exon creation and as well as the other issues. Notably, no sequencing reads covered that variant on the previous exome, which is to be expected because it is an exome. And this variance actually did not appear on the genome sequencing report that we had ordered through the UDN. In addition, the splice prediction tools like SpliceAI actually considered this variant to be unlikely to affect splicing. But with this, we felt confident in making the diagnosis of Renn-Penning syndrome. So I'll hand it back over to Carlos. So just a couple of comments. Renn-Penning syndrome is a condition that we all clinicians know, but it's relatively rare. And the clinical manifestations are relatively very non-specific. So intellectual disability, microcephaly, short stature and dysmorphic features are common, but this is something that we see in a lot of our patients. So the interesting thing is that when you look at the literature, these patients tend to be, even though they have broad impairment, they have some lean body build and scanty subcutaneous tissue. So it actually is some of the appearance that our patient has, often attributed to feeding difficulty. There are a number of articles that show that they have muscle atrophy that is progressive ankylosis of the MCP joints and biliard dysfunction. So in summary, this is our patient on the right. I think that once I go retrospectively and look at this diagnosis, I say, yes, it fits, but this is not the diagnosis that most clinicians will do without the help of the diagnosis. And on the left, there's an article by Germanow and Grube that shows multiple patients with Renn-Penning syndrome. And you can see that there's some interesting similarities with our patients. So we are happy that now we have a good diagnosis. So next, I think it's back to you guys. Thanks, Carlos. I'll go ahead and talk about our second case today. and this little girl, when we evaluated her application to the UDN, she was a seven-year-old female with intellectual disability, dysmorphic features, and seizures, and you can see a picture here of her that was taken when she was nine. She does have a history of hypotonia from infancy. Her medical history is also significant for asthma, laryngotracheal malacia, and recurrent ear and upper airway infections. She has high hyperopia and accommodative esotropia, and then her seizure history was starting with seizures around five years of age with focal motor onset. The seizures have been well controlled with Keppra. On physical exam, her growth parameters were within normal limits between the 75th and the 90th percentile. Again, there were some dysmorphic features which included, you can appreciate in the image here, hypertelarism with slightly upslanting palpebral fissures, telecanthus, mild epicanthal folds. She also has a thin vermilion but a fleshy lower lip, long philtrum, and you can really sort of appreciate it, a unique nose that's a broad nose with a tubular shape. Also possibly had macrodontia. She had wide incisors, and I'll show pictures of that in a little bit. Also had broad hallucines with soft palms and scoliosis. Developmentally, she did show delays. She sat at eight months, crawled at 19 months, pulled to a stand around 18 months, and walked at 21 months. I did mention she had intellectual disability. Her IQ was judged to be in the 50s. In terms of her speech ability, vocabulary included 20 to 30 words. She was combining two to three words, but her speech was somewhat indistinct. Her evaluations that she had had prior to coming to the UDN included a negative trio exome in 2015, which was then asked to be reanalyzed in 2017, and it was still negative, no new findings. She'd also had a history of a chromosomal microarray, and that was back in 2012, and that was normal as well. Here are some more images of her. You can see her over time as she gets older. You can also see her teeth. Again, we were thinking possibly wide incisors. Again, her growth was normal. As part of the UDN evaluation, we did include RNA sequencing, and I will turn things back over to Dr. Murdoch to talk about those results. Thank you, Jill. Yes, so for this case, of course, we did the RNA-seq directed analysis, and we found that she had significantly decreased expression in the KANSL1 gene, so about 50% normal level of expression. This gene is interesting because it is associated with Coulan-DeVries syndrome. We went back and we looked at both the exome data and did not find any variants there to explain that finding. Then we also went back and looked at the chromosomal microarray, and there was some evidence of a deletion at this locus. However, it was considered benign given that it exists in a structurally complex region where it's very difficult to call variants. There's a lot of common background CNV variation where other similar losses have been reported. But given the RNA-seq finding and the good phenotype match, we proceeded with some additional studies, including a PCR analysis. Actually, after having done genome sequencing through the UDN, which also came back negative, but with the PCR analysis, we identified a de novo 307KB heterozygous deletion in the proband that removed the first two exons of this gene, KANSL1. It was de novo, so it was not present in either parent. Interestingly, the same issues that had been encountered with the CMA led to the genome sequencing being unable to properly align in that region as well. It had failed to call a CNV there, but ultimately because of the RNA-seq and the good phenotype match, we sent this different study where we were able to confirm the de novo deletion that led to the removal of these first two exons and the subsequent decreased expression in KANSL1. With that, I will hand it again back over to Jill to speak more about Coulan-De Vries. Thanks, David. So just revisiting our diagnosis here. And so, sorry, the Coulan-De Vries syndrome, just listing some features that are common in the condition with the overlap with our patient underlined. You can see, you know, she had the delays, the hypotonia, her dysmorphic features, specifically speech and language delay as well as epilepsy. Coulan-De Vries may also be associated with other congenital abnormalities and features listed here that were not necessarily present in our patient. And then some images here taken from gene reviews of individuals with Coulan-De Vries, and you can appreciate there are some phenotypic similarities here. So again, you know, kind of underlying, underlining our sort of support and the belief that yes, you know, this does seem to be a good fit for her diagnosis. And so with that, I will turn it over to Dr. Chow to talk about our third case. Hi, everybody. So the third case that we wanted to share with you, sorry, is a five-year-old boy of Pakistani descent who was referred to the UDN, our team, because he had this, you know, history of very delayed development. He was essentially nonverbal, nonambulatory, but there was this discordance between his language understanding. So even though he couldn't speak, could only make vocalizations, he seemed to have some preserved level of understanding of simple commands, very low central tone, and then very pronounced lower extremity spasticity, very hypertonic. And he was overall short in stature, both height and weight all tracked less than third percentile, but he remained normal cephalic throughout records over the five years. So he never had any regression of his head growth and had progressively become more bradykinetic and was now largely just kind of culture dependent. Any minimal milestones he had achieved really stagnated or was further lost. And what was interesting was that his family history was very significant for consanguinity. There was multiple relatives with premature lethality at less than six months of age and due to some type of an undiagnosed neurodevelopmental disorder. So this is a family pedigree that we were able to obtain. And as you can see, four generations back from this family, there is relatives on both the paternal and maternal side who are siblings. And on the maternal side, in the grandparent generation, there were these four individuals all reported to have lethality, premature death at less than six months of age, reportedly never cried, never developed speech, never developed any really significant motor milestones, and the family had no indication of why this was the case. And then there's this relative on the paternal side which had reportedly some cognitive and developmental delay but was not as severe, of course, as these maternal relatives. And then our individual here had two healthy older siblings who were unaffected and developing well. But his particular presentation, as far as the family was aware, was unique to their family unit. So on a physical exam for him, what we noted was that he was brachycephalic. There was indication from the family that he was rather unique, that he was the only male on both the paternal and maternal side of the family who was red-headed, fairer than the other relatives. He was vocalizing, very sociable, smiling, making good eye contact, was interactive and playing, but very, very hypotonic. And he had a significant hypertonia in his lower extremities. He was very hyperreflexive, had developed ankle clonus. He also had some very mild facial dysmorphism. So we noted the very widely spaced teeth, the upturning on the nose, and a little bit smoother philtrum, but really overall didn't see any striking dysmorphisms beyond these that we noted. There have been a series of neuroimaging studies that have been done. The one at four and a half years of age, which revealed that overall there was no structural abnormalities. The white matter tracts here, as you can see in the corpus callosum, are very full, well developed. The supratentoral structures, the cortex itself was full. There was no signs of any atrophy or loss. Cerebellar structures looked well, as well as brain stem across the board. And when this was compared, sorry, this was compared to him older. He's actually older by the time we saw him. He also still didn't have any major changes in his brain structure. So there was no signs of any reduction in size or loss. The overall spinal cord findings were also unremarkable. They had done this to make sure that he didn't have any spinal cord tethering or other anatomic abnormalities that could explain the spastic paraplegia that they were seeing. The genetic testing that had been done for this child pre the evaluation at the UDN included Fragile X repeat testing, which was normal at 30 repeats. He had also been considered for spastic paraplegia type 22. And so SLC16A2 gene sequencing was sent and was unremarkable. We had a repeat chromosomal microarray that was done as part of the UDN evaluation that revealed one no known pathogenic copy number variance. But of course, consistent with the family history, there were multiple contiguous regions of copy neutral absence of heterozygosities that were greater than seven megabases across multiple chromosomes. He had a clinical trio exome sequencing that had been performed and was non-diagnostic. It did identify a heterozygous known pathogenic variant in XOSC3. This is a gene that is maternally, a gene that is associated with autosomal recessive pontocerebellar hypoplasia type 1B, and the variant was maternally inherited. But there was no other evidence of another mutation or copy number loss on the other wheel that could explain this as a potential diagnosis for him. And with that, I'm going to hand this over to David. Oh, I'm sorry. So the final thing that I was going to add was that we did have a strong panel of differential diagnoses that were considered for this child. But again, all of it was excluded either due to the presence of like preserved reflexes, the early congenital onset that made spinal cerebellar ataxia is a little bit less likely. The preserved head circumference made several of the diagnoses also less likely in his consideration for the differential. And thank you, David. I'll hand this over to you now. All right. Thank you, Twan. So this case is a bit different in that the causative variant was actually found by a candidate variant approach. So essentially, a very astute interpreter, exome interpreter, looked at the previous exome that had been done and noted that he was homozygous for a variant in the gene prune 1. And what's interesting about this variant is that it's a synonymous one. So it doesn't affect the amino acid, but it's located the very last nucleotide of exon 6 in this gene prune 1. And what's also notable is that this gene is within the region of homozygosity that they had detected on the previous SNP array. And ultimately, it was found that the parents are heterozygous carriers for this. And this variant, using a few different prediction tools, is suggested to affect splicing. So human splice finder predicted that it most likely would affect splicing. Interestingly, splice AI, again, at least for this particular variant, predicted that it would not. And then the CAD score was 14, which is reasonable, especially for a synonymous variant. So but again, this particular variant was detected not off of the RNA-seq directed approach, but was actually found on the reanalysis of the existing exome. And so prune 1 is associated with this condition, prune 1-related neurodevelopmental disorder that Tuan will speak about more in just a few minutes. But when given that initial finding, this synonymous variant that was predicted or likely to affect splicing, we then went and looked at the RNA-seq data and found that, yes, indeed, it did affect splicing. And it led to homozygous exon 6 skipping. And so again, you see here the proband is in red, and then there are two controls in blue and green. And you clearly see that exon 6 in the proband is completely absent. What's also notable about this exon 6 is that the deletion of it is in frame. So it would not be expected to lead to NMD, which is why we didn't see any effect on expression, gene expression. However, though this was originally identified via a candidate approach, when we then put this case through the RNA-seq directed pipeline, we found this as well. So we would have found this even if we hadn't done the research reanalysis. But it's very important to point out that sometimes even variants that are in the exome may not be reported. And this certainly was not present on the exome report, but it may just have been overlooked because it's a synonymous variant. And so I'll hand it back over to Twan. Thank you, David. Yes, so this is the PRUN1-related neurodevelopmental disorder. Its official nomenclature now is a neurodevelopmental disorder with microcephaly, hypotonia, the low muscle tone, and variable brain anomalies. And so what was unique is the cardinal features include what was listed. And in our individual, one thing that was different was that he did have maintenance of his head circumference. But he did have profound truncal hypotonia, peripheral spasticity. He was non-ambulatory, non-verbal. And what's interesting is that even in the initial five individuals who were identified in 2015 with this condition, two of them had very similar features as our proband, including preservation of their head circumference. And so there is some variability in the severity of the spectrum. What is interesting and what's been known currently in animal model studies is that the complete deletion of this gene is very deleterious to early embryonic development, essentially results in lethality due to disruption in the growth and vascular development in utero. And studies in vitro of the human variants who have been identified reveal that these prune one mutant proteins cause an alteration in microtubule dynamics. So then it has problems with how the neurons are able to proliferate or migrate. And so this is felt to underlie the microcephaly presentation. And then they can also develop multiple cortical and subcortical abnormalities. They can have reduction in the corpus callosum and other white matter tracks. So our thought in this individual, because the synonymous variant keeps the protein in frame, and so you are still having expression of the transcript, and it's probably resulting into an altered protein structure, but it is present, that that probably still preserves some function of prune one. And so that would explain the overall maintenance of the head circumference in this individual versus the more commonly seen microcephaly. However, he still does have quite significant intellectual and cognitive impairments due to other reasons associated with this mutant protein. So I will turn it back over to David. Thank you. Okay. So kind of in conclusion, at least some of the advantages of a transcriptome-directed approach are the following. No a priori information is needed, as it doesn't require a candidate to first be identified on the primary analysis. Number two, it's agnostic to presumed inheritance patterns that may be incorrectly assigned due to variable expressivity. This technique, it may point to variants below the resolution of other assays. For example, things like single exon deletions or mosaic variants that maybe just are not called. And it's also suited for novel gene discovery, and it doesn't require a trio if you have a large enough cohort to be able to detect those outliers. But of course, there are always some limitations that are more inherent to RNA-seq in general. You know, though we, in our case, in our cohort, we felt that fibroblasts were superior to blood, there are still genes that are not well expressed in clinically accessible tissues, or that may undergo tissue-specific alternative splicing. And, you know, this technique will tell you what happened, but it doesn't tell you why that happened, which is why you may find the defect in the RNA-seq, but then you have to go back and look for the cause in the exome or genome or CMA or whatever other study is necessary to find out why you saw that difference in expression or splicing. Sometimes, you know, the problem may actually not be in the RNA. It may be at the protein level, in which case maybe a different kind of assay would be more, would be diagnostic and not RNA-seq. I mentioned that some genes undergo alternative splicing. And one, I guess, limitation of this technique is that it does require a large control data set in order to be able to detect some of these outliers. And ideally, you would want to have somewhere around 60 cases from the same tissue. But some of this may be overcome by data sharing efforts that are starting to happen. So in conclusion, we feel that RNA-seq can be very effective for making diagnoses in neurodevelopmental disorders. We demonstrated that exome sequencing, CMA, and to some degree, genome sequencing, do not capture everything. And so some things are going to be missed. In our cohort, fibroblasts were better than blood for detecting clinically relevant genes. And we showed, especially with that Renn-Penning syndrome case, that it's difficult to interpret non-coding changes that we encounter on genome sequencing, even with the latest tools for predictions. And we felt that this is a good assay, in particular, to complement genome sequencing in the case of exome or CMA negative cases. And so these are just some references that you all can look up for some more information. And then specific to RNA-seq, including the first one on the list, which is our paper that recently is in press, describing this work in more detail. And here's some acknowledgments from Baylor and the referring physicians, as well as our colleagues that developed some of the pipelines that we've used, and then the UDN as a whole. And with that, I think we can open things up to questions. So I'll hand it off. Thanks, David. Going into questions from attendees, one common question, going back to case one, so for Renn-Penning syndrome, did we do RNA-seq in mom to see if she also showed the abnormal splicing? So I can answer that. The answer is no, we do not have the RNA-seq from mother. It's something we could do, but we felt that, given the findings in the proband and the good phenotype fit, that it wasn't necessary. But yes, we could have done that. So also related to the Renn-Penning case, do we have information about the ratio of normal expression to the misplaced? Could this be affecting the phenotype at all? Well, yes. I mean, that's certainly a possibility. I mean, what you saw there on that sashimi plot was that even though NMD was occurring due to that out-of-frame pseudo-exon creation, there still was some being transcribed. So it wasn't complete NMD. So there is some being produced, and it is likely that the dosage effect is possibly affecting the overall phenotype. So I don't have specific information about the ratios, but it is likely that there is some, I guess, effect on the overall phenotype given the level of alternative splicing that's occurring. And what about, have we done any functional analysis of the splicing variant, or what about the downstream intron retention that we saw? How do we interpret that? Yeah, that's really interesting. So what's also interesting, that gene, PQBP1, is actually an important splicing gene. So it actually plays a role in generalized splicing. So we're not exactly sure what to make of that more distal intron retention that was occurring, but it could possibly be due to the fact that PQBP1 does, in general, play a role in splicing. But there could also be some downstream effects from that deep intronic variant as well, more distally. So that's a good point. But we have not done any follow-up functional studies to investigate that further. Some questions have come up more generally about RNA sequencing. In terms of, what about the timing of doing this? Some labs recommend RNA-seq concurrently with the DNA-level sequencing, or do we do RNA sequencing after the DNA level? Is there some benefit to doing it sequentially? Yeah, those are great questions. And I think we're still in an early phase where we don't really know what the best answer is. From that last slide, our conclusion was that perhaps this is best as a complement to genome sequencing when you have already done exome and CMA and have not found anything, but your suspicion is very strong for a genetic condition. So perhaps, from my standpoint, I think genome plus RNA-seq together are very powerful and are more powerful than one or the other. Because the problem with genome sequencing is that it's just so difficult to interpret those non-coding variants. For a number of reasons, ACMG criteria are mainly applicable to coding changes. And our splice prediction tools are better than they used to be, but they're not without their faults. So my personal opinion is that genome plus RNA-seq together would be best. In particular, if your exome plus CMA or your other genetic studies have been negative. But it's always good to have them together, the genome and the RNA-seq together, rather than doing one first and then doing the other, because they complement each other. So do we have any idea about what the yield is for this type of approach after negative exome sequencing? Yes. In fact, we do. And so our paper does describe that. And so what we found was that in our cohort, excluding cases that had been solved by either reanalysis of the exome or clinical evaluation, in the remainder of cases where we did not have a diagnosis, despite sequencing and other assays, genetic evaluations, that by using this technique, we were able to diagnose an additional about 17% of our cases. And that was across multiple different phenotypes, the majority being neurologic, but also included other, some skeletal phenotypes and others. So yes, I mean, which is very, I think, very good. Our overall UDN solve rate is about 35%. So this is, on top of that, we were able to solve about another 17% using this strategy. There are some questions related to controls. For instance, can you tell us a little bit more about the control data set that's being used? And how many control samples do you need? But also, how do you identify in these calls what might be a disease-causing variant when you're comparing to controls versus normal background rate sort of noise variation? Sure. So yeah, I'll try to explain that as best as I can. So when we say controls, maybe that's a little bit not accurate in that we basically, we had about 115 probands, including, and then another 50 or so parents that we had RNA-seq data from. And so we kind of pooled everybody together. So that was kind of our large control data set. And then by using this pipeline called DROP, we then looked for outliers across the entire cohort. So with that, we were then able to detect subtle differences in expression or splicing that were unique to an individual. And so we focused on those and then went back and looked at either the exome data or the genome data or CMA to see if we could explain why we found that. So it's, but this technique did require that large data sets in order to detect those subtle outliers. And we estimated that you need anywhere from maybe like 50 to 60 cases in general, RNA-seq cases to be able to detect those differences in expression and splicing. So we had more than that. And so we were able to detect quite a few. And some of them are in potentially novel disease genes that we're investigating further. But I would say anywhere from maybe like 50 to 60 cases from the same tissue and ideally sequenced on the same platform to minimize the confounders is necessary to find those differences. So I don't know if that answers the question. There's also been several people interested in knowing about if this can be used clinically and what is the cost associated with doing RNA-seq? So, you know, I don't know that anyone is formally doing this clinically, you know, like in a capillary lab like that. It's mainly on a research basis, you know, that could potentially change as we go forward. The, you know, so everything is very much in the research space. So, you know, so, you know, the cost, you know, RNA-seq is not, you know, that expensive, but then itself, but then, you know, the analysis, the interpretation, all of that, you know, is additional time and cost. So it's difficult to really say how much it would be. But, you know, I think it would be probably in line with, you know, other genomic study costs. But, you know, I would imagine that in the coming years, this likely will become a clinical test because of the utility that we and others in the UDN have and outside the UDN have demonstrated. Going back to some of our case examples for the case number two, Kuhlen-DeVries syndrome, was the deletion that we ultimately found in the KNSL1 gene able to be seen on the genome, in the genome data? So in that case, no. The problem was that, you know, again, that region, that five prime region of the gene has, you know, a lot of, it's a very complex region. And so it's difficult for the reads to align to that area. And so it's, you know, the same problem that they had with the CMA was true in the genome as well. So the genome sequencing was not able to call the same deletion in that region. And so, but again, because of the RNA-Seq finding and the phenotype fit, we had to go to a different type of assay to confirm it and where we identified the two exon deletion at the five prime end. And for the third case, so for prune one, do we have any feel for how common it is to find these variants that might be slipping through that are at the intron-exon boundary that might be synonymous and not being flagged? I don't know of any good, you know, numbers. But, you know, we, you know, at Baylor, we always, if a patient has had an exome previously, we will do a research reanalysis of that exome. And in some cases like that one, we're able to find a causative variant. So, you know, I think it depends. You know, I wouldn't be surprised if, you know, some variants like synonymous variants or ones that are maybe just outside the canonical splice sites are being filtered out. But if, you know, if an astute, you know, interpreter reanalyzes the case, they can see, you know, that this variant likely is to affect splicing. But, you know, I wouldn't be surprised if that's happening more often than we might like, which is another reason why, you know, research reanalysis is very important. But, you know, and sometimes that's just not able to be done or, you know, again, you know, in the case of a genome, you know, you have so many variants. And so another, you know, an alternate technique like the RNA-seq directed approach might be beneficial so that you can really, you know, focus on particular genes of interest and look for the causative variant. And for that prune one case, can you explain why that change actually leads to skipping of that exon? Yeah, I mean, it's in that, you know, it's that last nucleotide in exon six and it, you know, abolishes the normal splicing of that exon. So it just kind of skips over that particular exon. So, you know, it's within that canonical splice site. So. So there's also an interesting question that's been raised. And what if our, this attempt to do RNA-seq and fibroblasts is normal and doesn't find an answer, especially thinking about neurodevelopmental phenotypes, is there a way to get to our, you know, genes that might only be expressed in neuronal tissues? Yes, I mean, that's a great question. So, you know, we did show that, we did show that, you know, fibroblasts are better than blood, you know, but that doesn't mean that they are the end-all and be-all. You know, for neuro-related genes, you know, nothing is better than having brain tissue. But, you know, it's very difficult to do a brain biopsy. So, you know, there is work to do like transdifferentiation, you know, converting fibroblasts into neurons. And so that's, you know, hopefully a strategy that can be used to, you know, not have to do brain biopsies, but still get expression of genes that are normally only expressed in that tissue. So, but yes, I mean, you know, those are strategies that are being investigated now, and hopefully will be, have results in, you know, in the near future. But I do think, you know, that's a big problem, you know, that there are still genes that even in fibroblasts, we're not going to be able to see just because of the low expression, and that may be leading to, you know, cases not being solved. Are we aware at all about ACMG-type criteria to be used to help with interpretation of RNA variants, like there are for DNA variants? I'm not aware of that. But, you know, it certainly would be helpful. And I would imagine that as this field progresses, that's something that will likely be done. Okay. So what about detecting mosaicism? Can RNA-seq detect, help us detect mosaicism? So it can, and if the variant that you find, or the mosaic variant, if it is causing some defect in the RNA-seq, whether expression or splicing, we likely would be able to detect that, and then that would allow you to go back and look at your exome, your genome panel, or whatever, and look at that gene, you know, maybe more carefully, and find if there is a variant at low allele frequency. So yes, I think that is one thing that this could help with. So I do know we're almost out of time here. One request, David, if you could go back to the reference screen. There's also, these slides are going to be part of what's available in the Academy, Genetics Academy. So there were some requests to see those, maybe the RNA-seq as well. And then, so I guess since we are almost out of time, I know we didn't get through all of the questions. But we do appreciate you for joining us. And remember to go ahead and complete the evaluation, and to help, you need to do that for the CME credits in particularly, and that's again through the Genetics Academy and ACMG. So thank you everybody for joining us. Thank you. Thank you, everyone.

RNA-seq

neurodevelopmental disorders

abnormal splicing

PQBP1 gene

Rett-Penning syndrome

decreased expression

KNSL1 gene

Coulan-DeVries syndrome

synonymous variant