This repository contains the implementation of the DCCNV pipeline for copy number variation (CNV) detection in single-cell DNA sequencing data. The pipeline includes data preprocessing, graph analysis, denoising using a diffusion process, contrastive learning, and segmentation using Circular Binary Segmentation (CBS).
DCCNV/
├── data/
│ └── A4_readcounts_processed.tsv
├── src/
│ ├── data_preprocessing.py
│ ├── graph_analysis.py
│ ├── denoising.py
│ ├── contrastive_learning.py
│ ├── segmentation.py
│ └── main.py
├── README.md
├── requirements.txt
├── LICENSE
└── .gitignore
The required Python packages can be installed using: pip install -r requirements.txt
To run the DCCNV pipeline, use the following command:
python src/main.py data/data.tsv
This project is licensed under the MIT License. See the LICENSE file for details.
The data_preprocessing.py module handles data loading and normalization. It includes the DataPreprocessor class with methods to load and normalize the data.
The graph_analysis.py module performs graph analysis including k-NN graph construction and affinity matrix computation. It includes the GraphAnalyzer class with methods to compute adaptive k, construct the k-NN graph, and compute the affinity matrix.
The denoising.py module implements the denoising technique using a multi-scale diffusion process. It includes the Denoiser class with methods to apply the diffusion process and denoise the data.
The contrastive_learning.py module defines and trains a contrastive learning model using a Siamese network architecture. It includes the ContrastiveLearningModel class with methods to create the base network, create the contrastive model, and compile and train the model.
The segmentation.py module performs segmentation using Circular Binary Segmentation (CBS). It includes the Segmenter class with methods to segment the input signal, convert the segmented result to an array, normalize the segmented result, and plot the segmented signal against the ground truth.
The main.py script runs the entire DCCNV pipeline. It includes data preprocessing, graph analysis, denoising, contrastive learning, and segmentation steps. The script also plots the segmented signal against the ground truth for comparison.
To run the pipeline with the provided example data, use the following command: python src/main.py data/data.tsv
python src/main.py data/A4_readcounts_processed.tsv