Validation of a Clinical Algorithm for the Diagnosis of Recessive Ataxias

Overview

The field of clinical diagnosis of recessive cerebellar ataxias (ARCA) is particularly complex and Next Generation Sequencing (NGS) techniques have revolutionized this neuro-genetic field. The current challenge is to optimize the analysis of genetic data generated by NGS because: the processing of data remains very laborious; diagnostic yeld less than 50%; the interpretation of the variants sometimes very difficult. For this purpose of optimization, the team of the University Hospital of Strasbourg has developed a computer algorithm based on 124 clinical and para-clinical parameters (derived from the data of the literature), useful to guide the genes to be targeted in priority by genetic analysis, in the context of a suspicion of ARCA (\> 60 known genes); this algorithm was validated retrospectively in 834 patients with genetically confirmed ARCA (92% Sense, 95% Spec). However, these 834 patients are often the same as those described in the literature and used for the elaboration of the algorithm. This introduces a bias in the initial evaluation of the algorithm, which therefore requires validation in clinical practice, from a cohort of patients referred for suspected ARCA (with or without a found genetic mutation). At the same time, Montpellier's genetics laboratory has developed a bioinformatic method for the search for copy number variations (CNV) that can be applied in a targeted manner to the genes predicted by the algorithm. The principal aim of this study is the validation of a semi-automated clinical algorithm for NGS molecular diagnosis of ARCA; the secondary objective is to evaluate if the application of this algorithm coupled with a targeted bioinformatic analysis can increase the diagnostic yield of the NGS analysis.

Design: Retrospective study based on clinico-genetic data. Total research duration: 22 months Plan of the study : More than 150 patients referred for ARCA suspicion have been analyzed by NGS (Montpellier platform) since September 2013. The clinical data of these patients will be entered at the computer level into the algorithm to obtain the prediction of the gene involved (under the form of a probability for each of the genes known to ARCA). The investigators will compare this result with that obtained by standard NGS analysis. For patients without molecular diagnosis defined after standard NGS analysis, the investigators will take the 5 most probable genes selected on the basis of the algorithm in order to carry out an in-depth analysis of the variants found and a bioinformatic analysis by semi-automatic detection of CNVs. Main Evaluation Criterion: Concordance of the algorithm with standard NGS analysis for the genetic diagnosis of ARCA. Secondary evaluation criterion:% of patients for whom the prediction based on the algorithm suggested the correct diagnosis (subsequently confirmed after revision in the detail of the genetic data) while the standard genetic analysis was not not informative. Statistical Analysis: The investigators will perform a concordance analysis (Cohen's k) to validate the algorithm in clinical practice. For a given patient, there will be concordance if: i) one of the first 5 genes predicted with the highest probability by the algorithm is compatible with the mutation found after standard NGS or ii) if no gene is predicted by the algorithm (score \<20) and no mutation found by NGS analysis.