Published on
In a publication in the journal Genetics in Medicine (Nature group), these teams show that this "automated diagnosis" based on unstructured biomedical data opens up new possibilities for the identification and diagnosis of patients with rare or common genetic diseases.
Discovery of a common mutation of the KCNA2 gene
Advances in diagnostic technologies, genetic testing and DNA sequencing, have led to the discovery of new genes and diagnostics for a wide range of rare diseases, including epilepsy. However, the heterogeneity of genotypes (genetic characteristics of an individual) and phenotypes (physical and biological characteristics of an individual) makes diagnosis complex in the majority of cases. The KCNA2 gene, when altered, can cause symptoms ranging from isolated intellectual disability to developmental and epileptic encephalopathies, in a phenotypic spectrum that can make its diagnosis complex.
The reference center for rare epilepsy, led by Rima Nabbout at the Hôpital Necker-Enfant malades AP-HP, the translational research for neurological disorders team, led by Edor Kabashi at Imagine, the teams of the department of pediatric neurology at the Hôpital Necker-Enfant malades AP-HP, with the Dr. Warehouse software, have developed the multimodal deep phenotyping of two patients with the same KCNA2 gene mutation, i.e. the detailed description of their clinical and biological status through different types of data, such as neonatal history, description of movement disorders or seizures, neurological examinations, clinical course of developmental milestones, feeding, sleep disorders, metabolic or genetic screening, or any relevant medical conditions.
Dr. Warehouse, detecting patients by similarity using artificial intelligence
These patients could be identified and compared thanks to the Dr Warehouse program, developed by Nicolas Garcelon and Codoc, an Imagine spin-off resulting from an ongoing collaboration between the Imagine Institute and the Hôpital Necker-Enfants malades AP-HP. Dr. Warehouse automatically analyzes and extracts phenotypes from approximately 500,000 patient records, representing more than 6 million unstructured documents, such as hospitalization reports, medical reports, discharge letters, imaging reports, biological results. A similarity query allows to find similar patients from their automatically extracted phenotypic characteristics and to establish a specific phenotypic profile.
The use of information systems capable of processing texts and documents is particularly suited to centers specializing in the diagnosis and treatment of rare diseases, for which rare or new symptoms or signs may be absent from structured data. For example, this publication highlighted that in the absence of structured data, Dr Warehouse revealed the relevance of symptoms such as sleep disorders and GERD (gastroesophageal reflux disease) in the disease caused by the KCNA2 gene mutation.
This publication shows the importance of in-depth phenotyping and paves the way for the use of unstructured data systems in the field of rare diseases for diagnosis, cohort extension, and examination of the natural history of rare and common diseases.
Deep phenotyping unstructured data mining in an extensive pediatric database to unravel a common KCNA2 variant in neurodevelopmental syndromes
Marie Hully, Tommaso Lo Barco, Anna Kaminska, Giulia Barcia, Claude Cances, Cyril Mignot, Isabelle Desguerre, Nicolas Garcelon, Edor Kabashi & Rima Nabbout
Genetics in Medicine (2021)
To read the publication : https://www.nature.com/articles/s41436-020-01039-z