Mots-Clés
AI for health
longitudinal data analysis
rare disease data science
Description
Master 2 Internship Project – Bioinformatics / Biostatistics / AI for Health
**Quantitative reconstruction of the natural history of gamma-sarcoglycanopathy using retrospective clinical data****
Gamma-sarcoglycanopathy, also known as LGMD R5, is a rare autosomal recessive limb-girdle muscular dystrophy caused by pathogenic variants in the SGCG gene. It is one of the most severe forms of sarcoglycanopathy, typically beginning in childhood and leading to progressive proximal muscle weakness, elevated creatine kinase levels, loss of ambulation often before adulthood, and variable cardiac and respiratory involvement. Despite available clinical, functional, genetic, histological and imaging data, the natural history of LGMD R5 remains insufficiently characterized from a quantitative and predictive perspective. This represents a major limitation for patient stratification and for the design of future therapeutic trials, particularly in the context of ongoing gene therapy development.
The aim of this Master 2 internship is to develop and apply data-driven approaches to reconstruct the natural history of LGMD R5 from retrospective longitudinal data. The project will rely on published and multicentre datasets including clinical variables, functional scores, age at disease onset, age at loss of ambulation, respiratory and cardiac assessments, muscle MRI data, biopsy findings and SGCG genotype information. The student will contribute to the curation, harmonization and analysis of these heterogeneous datasets, with the goal of modelling disease progression and identifying clinically meaningful trajectories.
The first objective will be to describe disease progression over time using longitudinal statistical models. Mixed-effects models, and potentially Bayesian hierarchical models, will be used to estimate average disease trajectories while accounting for inter-individual variability. These models will help quantify the natural course of LGMD R5 and evaluate the influence of factors such as genotype, age at onset, baseline ambulatory status or residual sarcoglycan expression on the rate of progression.
The second objective will be to identify subgroups of patients with distinct progression profiles. Unsupervised approaches such as hierarchical clustering, k-means, Gaussian mixture models or longitudinal latent class models may be applied to variables describing functional decline, muscle imaging progression or key clinical milestones. These analyses should allow the identification of patients with rapid, intermediate or slower disease progression and help characterize the clinical, genetic and biological features associated with each subgroup.
The third objective will be to explore predictive models of future disease evolution from early clinical data. Depending on data availability, survival models, mixed predictive models or interpretable machine-learning approaches such as random forests or gradient boosting may be used to estimate the risk of major clinical events, including loss of ambulation or cardio-respiratory involvement. Particular attention will be paid to model interpretability, uncertainty estimation and robustness, given the constraints of rare disease datasets.
The intern will be integrated into a multidisciplinary research environment combining expertise in neuromuscular diseases, translational research, bioinformatics, statistics and artificial intelligence. The expected work includes literature review, dataset structuring, data cleaning, exploratory analyses, implementation of statistical and machine-learning pipelines in Python and/or R, interpretation of results in close interaction with clinicians and researchers, and preparation of a written scientific report. The project may also contribute to the preparation of a scientific publication.
This internship is particularly suited for a Master 2 student in bioinformatics, biostatistics, data science, AI for health or computational biology, with strong interest in rare diseases and translational biomedical research. Skills in Python or R are required, and prior experience with longitudinal data analysis, machine learning, survival analysis or clinical data would be an advantage. The project will provide training in the analysis of real-world rare disease datasets and contribute to the development of quantitative tools supporting future clinical trials in gamma-sarcoglycanopathy.