A recessive ataxia diagnosis algorithm for the next-generation sequencing era

Abstract

Objective: Differential diagnosis of autosomal recessive cerebellar ataxias can be challenging. A ranking algorithm that predicts the molecular diagnosis based on the clinical phenotype of a patient has been developed to guide genetic testing and to align genetic findings with the clinical context.

Methods: An algorithm that follows clinical practice, including patient history, clinical, MRI, electromyography and biomarker features, was developed following a review of the literature on 67 autosomal recessive cerebellar ataxias and personal clinical experience. Frequency and specificity of each feature were defined for each autosomal recessive cerebellar ataxia, and corresponding prediction scores assigned. Clinical and paraclinical features of patients are entered into the algorithm, and a patient's total score for each autosomal recessive cerebellar ataxia is calculated, producing a ranking of possible diagnoses. Sensitivity and specificity of the algorithm were assessed by blinded analysis of a multinational cohort of 834 patients with molecularly confirmed autosomal recessive cerebellar ataxia. The performance of the algorithm was assessed versus a blinded panel of autosomal recessive cerebellar ataxia experts.

Results: The correct diagnosis was ranked within the top 3 highest-scoring diagnoses at a sensitivity or specificity of >90% for 84% and 91% of the evaluated genes, respectively. Mean sensitivity and specificity of the top 3 highest-scoring diagnoses were 92% and 95%, respectively. The algorithm outperformed the panel of ataxia experts (P=0.001).

Interpretation: Our algorithm is highly sensitive and specific, accurately predicting the underlying molecular diagnoses of autosomal recessive cerebellar ataxias, thereby guiding targeted sequencing or facilitating interpretation of next-generation sequencing data. This article is protected by copyright. All rights reserved.

http://ift.tt/2h1JuPm