Disease Progression Score Estimation From Multimodal Imaging and MicroRNA Data Using Supervised Variational...

Area of Science:

• Neuroscience
• Biomarker Discovery
• Computational Biology

Background:

• Frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS) are rare, progressive neurodegenerative diseases lacking effective treatments.
• Accurate assessment of disease progression is critical for developing and evaluating new therapeutic strategies.
• Current biomarkers, including neuroimaging and microRNA data, show promise but lack the precision to individually track FTD and ALS progression, especially given the challenge of small sample sizes in rare diseases.

Purpose of the Study:

• To develop a novel computational framework for calculating a disease progression score (DPS) using multimodal data from small, cross-sectional samples.
• To create a method that integrates various biomarkers to more accurately model disease trajectory in rare neurodegenerative conditions.
• To establish an objective measure for disease progression that can aid in the clinical trial evaluation of novel therapies for FTD and ALS.

Main Methods:

• A supervised multimodal variational autoencoder was developed to infer a latent space representing disease progression.
• Latent representations were aligned along a disease trajectory, and a DPS was computed via orthogonal projections.
• The framework was validated using synthetic datasets and a real-world dataset from the PREV-DEMALS study, comprising patients, presymptomatic carriers, and controls.

Main Results:

• Performance evaluation on synthetic data confirmed the utility of the area under the ROC curve (AUC) as a proxy metric for DPS accuracy.
• Experiments on the PREV-DEMALS dataset demonstrated superior performance of the proposed framework compared to existing state-of-the-art methods.
• The framework successfully leveraged small, multimodal datasets to objectively quantify disease progression.

Conclusions:

• The proposed framework offers a robust method for computing a disease progression score (DPS) from limited, multimodal data in rare neurodegenerative diseases.
• This approach has significant potential for application in clinical trials by providing a more objective measure of disease progression.
• The ability to combine multiple biomarkers advances the objective assessment of progression in complex conditions like FTD and ALS, paving the way for better therapeutic development.