Revealing chronic disease progression patterns using Gaussian process for stage inference.

Area of Science:

• Computational biology
• Biostatistics
• Medical informatics

Background:

• Chronic diseases often progress slowly with subtle early symptoms, making modeling difficult.
• Disease heterogeneity and discrete patient observations complicate understanding progression.
• Accurate disease modeling impacts clinical practice and drug development.

Purpose of the Study:

• To develop a novel approach for uncovering chronic disease progression patterns.
• To assess the dynamic contribution of clinical features in disease progression.
• To stratify patient data for osteoarthritis, bipolar disorder, and hepatocellular carcinoma.

Main Methods:

• Developed the Gaussian Process for Stage Inference (GPSI) approach.
• Applied GPSI to synthetic and real-world datasets (OA, BP, HCC).
• Utilized unsupervised learning to disentangle temporal and phenotypic heterogeneity.

Main Results:

• GPSI identified distinct OA subgroups linked to genotypes and pathways.
• GPSI revealed two BP developmental patterns and brain atrophy contributions.
• HCC progression patterns were consistently reproduced across independent datasets.

Conclusions:

• Unsupervised methods can identify disease subgroups with common progression patterns.
• GPSI effectively disentangles disease heterogeneity.
• Findings enable personalized treatment plans based on disease stage and features.