Single-cell transcriptomic integrated with machine learning reveals human retinal cell-specific biomarkers in...

Area of Science:

• Genomics and Molecular Biology
• Ophthalmology and Vision Science
• Computational Biology and Bioinformatics

Background:

• Diabetic retinopathy (DR) is a major cause of vision loss globally.
• The precise cell-type-specific molecular changes driving DR progression are not fully understood.
• Identifying these changes is crucial for developing effective treatments.

Purpose of the Study:

• To create a single-cell transcriptomic atlas of the Chinese diabetic retina.
• To characterize transcriptional shifts in individual retinal cell types during diabetes and DR.
• To identify candidate biomarkers associated with DR pathogenesis using machine learning.

Main Methods:

• Generated a single-cell RNA sequencing (scRNA-seq) dataset from 297,121 retinal cells of Chinese donors (non-diabetic, diabetes, DR).
• Performed rigorous quality control, batch correction, clustering, and cell-type annotation.
• Utilized differential expression analysis and a machine-learning framework (L1-regularized logistic regression, RFE-CV, stability selection) to identify biomarkers.

Main Results:

• Identified 10 major retinal cell populations with extensive cell-type-specific transcriptional alterations in diabetes and DR.
• Pathway analysis revealed significant immune activation, oxidative stress, neurodegeneration, and synaptic dysfunction across cell types.
• Discovered 707 cell-type-specific candidate marker genes, offering a resource for disease mechanism research and therapeutic target identification.

Conclusions:

• Established a valuable single-cell transcriptomic atlas of the Chinese diabetic retina.
• Revealed distinct molecular signatures in specific retinal cells linked to DR progression.
• Provided critical biological insights and nominated potential biomarkers for future translational research in DR.