Project
MobiDeep
AI meta-score developed with the MoBiDiC group (CHU Montpellier) for prioritizing non-coding variants in whole-genome sequencing.
MobiDeep is a project focused on prioritizing non-coding variants from whole-genome sequencing data. The underlying problem is central in rare diseases: non-coding regions represent a large part of the genome, remain difficult to interpret and may contribute to diagnostic delay.
The project was carried out in collaboration with the MoBiDiC group (Montpellier Bioinformatics for Clinical Diagnosis), under the supervision of David BAUX.
Intent
The project combines benchmarking of existing prediction tools, calibration by genomic region and development of an artificial intelligence meta-score. The goal is to improve prioritization of hard-to-interpret variants while keeping a critical view of performance, limits and clinical context.
Public Sources
- Preprint: MobiDeep: an AI-based meta-score for scoring non-coding DNA variations
- Fonds Guilhem: Research project to accelerate rare disease diagnosis
- Fondation Groupama: Plongee dans l’ADN invisible
- Video: MobiDeep A Benchmark Driven Ensemble Learning Approach