scVIP: personalized modeling of single-cell transcriptomes for developmental and disease phenotypes.

Area of Science:

• Genomics
• Computational Biology
• Neuroscience

Background:

• Single-cell RNA sequencing (scRNA-seq) provides deep insights into cellular heterogeneity.
• However, connecting cell states to individual-level phenotypes is a significant challenge in biological research.

Purpose of the Study:

• To introduce scVIP, a generative framework designed to integrate transcriptional profiles and phenotypic markers.
• To enable the learning of personalized, individual-level embeddings for biological data analysis.

Main Methods:

• Utilized generative models and cell-type-aware multi-instance learning.
• Developed a framework to integrate scRNA-seq data with phenotypic markers for personalized embeddings.
• Implemented a method to harmonize datasets with varying phenotype definitions.

Main Results:

• scVIP successfully predicts developmental age, disease progression, and neuropathology.
• The framework harmonizes datasets with distinct phenotype definitions, enabling cross-study comparisons.
• Identified disease-relevant cell populations and key transcriptional programs associated with neurodegeneration.

Conclusions:

• scVIP offers a powerful approach to bridge the gap between cellular heterogeneity and individual phenotypes.
• The framework enhances the predictive power of scRNA-seq data for clinical applications.
• Provides novel insights into the cellular and molecular mechanisms of neurodegeneration.