scVI-3D

Overview: scVI-3D is a deep generative modeling framework designed for analyzing single-cell Hi-C data. It systematically addresses key challenges in 3D genome structural analysis, including band bias, sequencing depth effects, zero inflation, sparsity impact, and batch effects in scHi-C data.

Key Features:

1. Deep Generative Modeling:

   • Utilizes variational autoencoders for non-linear dimension reduction
   • Implements parametric count models (Poisson and Negative Binomial)
   • Captures complex patterns in chromatin conformation data

2. Comprehensive Data Processing:

   • Handles band bias correction
   • Addresses sequencing depth normalization
   • Manages zero inflation and sparsity
   • Removes batch effects

3. Scalable Analysis Pipeline:

   • Supports large-scale scHi-C datasets
   • Enables parallel processing
   • GPU acceleration support
   • Efficient memory management

4. Visualization and Analysis Tools:

   • UMAP and t-SNE visualizations
   • Principal component analysis
   • Batch effect correction validation
   • Cell type clustering analysis

Applications: scVI-3D has been successfully applied to analyze single-cell Hi-C datasets, enabling researchers to:

• Identify cell-type specific 3D genome organizations
• Study chromatin conformation heterogeneity
• Integrate multiple batches of scHi-C data
• Explore nuclear processes at single-cell resolution

The package is implemented in Python with comprehensive documentation and tutorials to support researchers in the fields of genomics and single-cell biology.

Visit our manuscript.