Author: Edgard Verdura, PhD
Welcome to our October edition of Molecular Mechanisms! This month we continue our Molecular Mechanisms series by diving into a case that illustrates how an accurate clinical description of a patient’s case can help the machine-learning prioritization of variants. Read our first edition here.
In the second example, a variant analyst from a well-known diagnostic service company uploaded the case of an adult patient through AION. The patient originally underwent medical evaluation due to deteriorating kidney function and was referred for genetic testing without any further clinical information. Genetic screening was performed using a nephropathy gene panel, combined with AION phenotype-based prioritization (using only “Nephropathy” HPO term as input). Through AION, the variant analyst found the diagnostic variant p.Cys508Arg (c.1522T>C, exon 7) in PKD1 gene, whose mutations have been linked to Polycystic Kidney Disease 1.
Although this variant was described as Likely Pathogenic in ClinVar, and accordingly predicted as Pathogenic by our ML score (with 90% confidence), initially, it was ranked at #19.
After reexamining the case, the user realized that the low ranking could be related to the poor clinical characterization of the patient when she submitted the case to AION. After confirming that this patient indeed was affected by Polycystic Kidney Disease, we decided to process the same case using the more specific HPO term “Multicystic Kidney disease", as initially suggested by AION when uploading HPOs for this case. After this HPO refinement, even if little, AION ranked the variant in the top position.
Therefore, this case serves as an example of how an accurate phenotypic description can help to provide a better ranking and, consequently, shorten the diagnostic odyssey of many patients.
PKD1 mutations are well known to be associated with Polycystic Kidney Disease 1, an autosomal dominant kidney disease characterized by the adult onset development of cysts in both kidneys, liver, and often also other organs, together with an increased risk of intracranial aneurysms (Harris et al., 2018). Interestingly, the PKD1 gene is also known for being especially difficult to sequence. Given its homology with six pseudogenes (highly similar regions of the genome), false positive variants can be generated due to parallel amplification of these pseudogenes, or also due to inefficient variant calling. To solve this issue, laborious long-range PCR methods were used in the past to study this gene. In the last years, targeted NGS sequencing methods that circumvented these difficulties were developed (Trujillano et al., 2014) and the recent development of long-read sequencing methods has helped to thoroughly screen this region without any interference of PKD1 homologous sequences (Borràs et al., 2017).
In upcoming posts, we will dive deeper into the importance of performing a correct variant calling. We will show you how to avoid the presence of contaminating false positive variants during tertiary analysis, which might hinder variant prioritization. Please stay tuned for more examples of cases solved with AION!