Add new README

README.txt

+
+# Genomic CNV and SV Detection with GPU Acceleration
+
+This project performs copy number variation (CNV) and structural variant (SV) detection on genomic data, leveraging 
+GPU acceleration to enhance performance for large datasets. It includes calculations of mappability, GC content, depth, 
+and normalization, followed by variant detection and result output in various formats, including VCF for SVs.
+
+## Features
+
+- **Copy Number Variation (CNV) Analysis**: Analyzes depth of coverage across genomic windows to detect CNVs.
+- **Structural Variant (SV) Detection**: Identifies SVs (e.g., deletions, inversions, translocations) using paired-end 
+  and split-read alignments.
+- **GPU Acceleration**: Utilizes CUDA-enabled GPU processing to improve the efficiency of mappability, GC content, 
+  depth, and normalization calculations.
+- **Customizable Parameters**: Adjustable settings for window size, step size, and z-score thresholds.
+
+### Author:
+**SERRALTA Theo**
+
+### Collaborators:
+**DUFFOURD Yannis**
+
+### Laboratory:
+**GAD** 
+
+### Date:
+**28/09/2023**
+
+## Installation
+
+1. Ensure you have Python and CUDA installed.
+2. Install the necessary Python packages:
+
+   ```bash
+   pip install numpy pysam pycuda pandas
+   ```
+
+3. Clone this repository.
+
+## Usage
+
+### Platform
+Currently, this software is designed to run exclusively on the CCUB (Computing Center of the University of Burgundy).
+
+### Directory
+Navigate to the directory:
+
+```bash
+cd /work/gad/shared/analyse/test/cnvGPU/test_scalability/
+```
+
+### Recommended Execution with qsub
+
+Run the following command to execute using qsub:
+
+```bash
+qsub -v NUM_CHR=<ALL_or_num_chr>,INPUTFILE=</path/to/the/input/bam/file>,LOGFILE=</path/to/the/log/file>,OUTPUT=</path/to/the/output/file>,OUTPUT_PAIRS=</path/to/the/output_pairs/file>,OUTPUT_SPLITS=</path/to/the/output_splits/file> ./wrapper_cnvGPU.sh
+```
+
+Example:
+
+```bash
+qsub -pe smp 1 -v NUM_CHR=ALL,INPUTFILE=/work/gad/shared/analyse/test/cnvGPU/test_scalability/dijen1000.bam,OUTPUT=exemple.out.tsv,OUTPUT_PAIRS=exemple.out_pairs.tsv,OUTPUT_SPLITS=exemple.out_splits.tsv,LOGFILE=exemple.log ./wrappers/wrapper_cnvGPU.sh
+```
+
+### Modifying Parameters
+
+Certain parameters can be customized within the wrapper script:
+
+- `window_size` (w): Default is `-w 100`
+- `step_size` (s): Default is `-s 10`
+- `zscore_threshold` (z): Default is `-z 1.5`
+- `lengthFilter` (l): Default is `-l 200`
+
+### Direct Execution without Wrapper
+
+Alternatively, you can execute the program directly with Singularity:
+
+```bash
+singularity exec --nv -e /work/gad/shared/bin/singularity_images/pycuda/pycuda_sam.1.1.sif python3 /work/gad/shared/analyse/test/cnvGPU/test_scalability/cnv_sv_caller_gpu.py -b <input_bamfile> -c <int or "ALL"> -w <int> -s <int> -z <float> -l <int> -o <output_cnv_file_vcf> -p <output_pairs_file> -m <output_splits_file> -e <logfile>
+```
+
+Example:
+
+```bash
+singularity exec --nv -e /work/gad/shared/bin/singularity_images/pycuda/pycuda_sam.1.1.sif python3 /work/gad/shared/analyse/test/cnvGPU/test_scalability/cnv_sv_caller_gpu.py -b example.bam -c ALL -w 100 -s 10 -z 1.5 -l 200 -o example_cnv.vcf -p example_pairs.tsv -m example_splits.tsv -e example.log
+```
+
+## Output Files
+
+- **VCF File**: Contains structural variant calls with relevant information on chromosome, position, variant type, 
+  copy number, etc.
+- **Paired-Read Events**: Details abnormal paired-end read alignments indicating possible structural variations.
+- **Split-Read Events**: Lists split-read alignments for further variant investigation.
+
+## Dependencies
+
+- Python 3.x
+- CUDA-compatible GPU
+- [Numpy](https://numpy.org/), [pysam](https://pysam.readthedocs.io/), [pycuda](https://documen.tician.de/pycuda/), 
+  [pandas](https://pandas.pydata.org/)
+
+## License
+
+This project is licensed under the MIT License.
+
+## Acknowledgments
+
+This tool was developed to assist with high-performance genomic analyses, utilizing GPU acceleration to make 
+large-scale CNV and SV detection feasible on big datasets.