A NCV implementation by Luyu

Installation

1. Install BWA and samtools
2. Prepare genome file hg38.fa
   1. Download wget http://hgdownload.soe.ucsc.edu/goldenPath/hg38/bigZips/hg38.fa.gz
   2. Unzip gunzip hg38.fa.gz
   3. Build index bwa index hg38.fa
3. Install python2/python3
4. Install numpy, pandas, pysam, tqdm, pp by pip
5. Download all files In this repository

Folder structure

1. NCV_ver_1.6.py: main executable python script.
2. genome_browser.py: auxiliary file to generate chromosome view html.
3. batch3_females and batch3_males: example of the format for -f/-m/-s arguments.
4. ncv.model: a pre-trained model to quick start.
5. hg38_cytoBand.txt: the hg38 reference cytoband annotation file.

Usage

1. use BWA to map raw reads to reference genome and sort them by samtools bwa mem -t 20 hg38.fa sample.fq.gz | samtools sort -@ 20 - -o sample.bam
2. index BAM fiels samtools index sample.bam
3. train or test
   1. train the model with Normal NIPT samples: python NCV_1.5.py train -m male.lst -f female.lst
   2. test a batch of NIPT samples: python NCV_1.5.py test -s test.lst

Output

The train model only outputs a model file specified by -m option (by default, ncv.model in current folder)

The test model will create a new folder specified by -o (by default, current date and time)。In that folder:

1. table.csv: A z-score excel table where row=sample and column=chromosome
2. table.html: A z-score web table where row=sample and column=chromosome
3. Subfolders: One subfolder for each samples. In each subfolder, every chomosome has one html file to view bin-based z-score, for manual inspection of micro-insertion/deletion.

---

Type python NCV_ver_1.6.py --help to show help information:

usage: NCV_ver_1.6.py [-h] [-p CPUS] [-d MODEL] [-m MALE] [-f FEMALE]
                      [-s TEST] [-o OUTPUT]
                      mode

A Python implementation of NCV for NIPT. 

positional arguments:
  mode                  <train/test> In train mode, -m and -f are required. In
                        test mode, -s is required. [required, default=test]

optional arguments:
  -h, --help            show this help message and exit
  -p CPUS, --cpus CPUS  Number of CPUs to use for training[default=20]
  -d MODEL, --model MODEL
                        The model file to be write in training or read in
                        test. [default=ncv.model]
  -m MALE, --male MALE  A file that each line contains a male training sample
                        path.
  -f FEMALE, --female FEMALE
                        A file that each line contains a female training
                        sample path.
  -s TEST, --test TEST  A file that each line contains a test sample path.
  -o OUTPUT, --output OUTPUT
                        Output filename prefix [default=<current time>]