Title: Computational analysis of 10,860 phenotypic annotations in individuals with SCN2A-related disorders
Affiliation: University of Pennsylvania, Children’s Hospital of Philadelphia, The College of New Jersey, Newcastle University, Royal Victoria Infirmary, University of Luxembourg
Authors: Katherine Crawford, Julie Xian, Katherine L. Helbig, Peter D. Galer, Shridhar Parthasarathy, David Lewis-Smith, Michael C. Kaufman, Eryn Fitch, Shiva Ganesan, Margaret O’Brien, Veronica Codoni, Colin A. Ellis, Laura J. Conway, Deanne Taylor, Roland Krause and Ingo Helbig
Reference: Genetics in Medicine 2021;23(7):1263-1272
We thank Dr David Cunnington for providing us with a summary of this work.
Summary:
What did they do?
For this study, the researchers collated data on 413 individuals with SCN2A-related disorders from publications on people with SCN2A between 2001 and 2019. For each of these individuals, details on clinical symptoms were coded in such a way that they could be digitised. This resulted in 10,860 symptoms across the 413 individuals. This data was then analysed using computer analysis to identify clusters of symptoms that mapped to distinct clinical presentations.
What did they find?
The authors grouped individuals in to 5 broad clinical categories (phenotypes): DEE, autism spectrum disorder, benign familial neonatal-infantile seizures, other epilepsies and atypical presentations. They then mapped the 5 most common symptoms associated with each clinical category (shown in the table). Then using these symptoms, they tested how accurately these symptoms were at predicting clinical presentations and variant function. This approach allowed correctly predicting gain of function versus loss of function variants in 84% of cases.
What does this mean?
Outlining the common symptoms of the main clinical presentations of SCN2A, allows those managing people with SCN2A-related disorders to group patients in to particular sub-types. This is helpful, particularly with the development of precision-based treatments, which will benefit different sub-groups of people with SCN2A. However, this technique is not perfect, with only accurately classifying 84% of those with SCN2A-related disorders as gain versus loss of function.
How does this help move the field forward?
This study shows the strength of adding computational analysis to the data collected clinically such as in natural history studies. It also highlights that each type of study is needed to get the best results. Careful and accurate clinical assessment, paired with strategies to digitise that data for computational analysis to find patterns over and above what can be recognised even by experienced clinicians.
