ChrNet

ChrNet predicts immune cell types from single-cell RNA sequencing (scRNA-seq) data by applying a chromosome-based one-dimensional convolutional neural network that incorporates gene positional information aligned to chromosomes.

Key Features:

• Chromosome-specific gene positional encoding: Incorporates gene positional data aligned to chromosomes as input context for classification.
• One-dimensional convolutional neural network (1D-CNN): Uses a chromosome-based 1D-CNN architecture to learn spatial patterns along chromosomes.
• Re-trainable supervised learning: Implements a supervised, re-trainable classifier to adapt to labeled datasets.
• Supports scRNA-seq data: Designed to process single-cell RNA sequencing (scRNA-seq) datasets for cell type prediction.
• Improved classification performance: Demonstrates benchmarked accuracy of over 90% in immune cell type profiling.
• Addresses unsupervised clustering variability: Provides a consistent classification alternative to unsupervised clustering methods that can yield variable results depending on input parameters and dataset size.

Scientific Applications:

• Immune cell type profiling: Classifies immune cell types from scRNA-seq data for single-cell studies.
• Tumor microenvironment analysis: Profiles immune composition within tumor microenvironments to inform biological interpretation.
• Cancer research and clinical inference: Supports analyses relevant to cancer prognosis and therapeutic strategy by improving immune cell identification in cancer tissues.

Methodology:

Incorporates gene positional data aligned to chromosomes into inputs and applies a chromosome-based 1D-CNN as a re-trainable supervised classifier on scRNA-seq datasets.