FSDA: Fragment Size Distribution Analysis for non-invasive prenatal CNV prediction

Requirements

• Python 2.6 or newer
• NumPy & SciPy
• samtools
• bcftools

Pre-process

Use the preprocess.sh script in the preprocess directory to create the required files for prediction. Based on the size of the sample and reference, this can take hours.

Usage:

./preprocess.sh [sample_dir] [ref_dir] [prep_dir] [ref_genome] [dbsnp_file]

• sample_dir is the directory containing the .bam and the .bai files for the sample cfDNA data
• ref_dir is the directory containing the .bam and the .bai files for the reference cfDNA data
• prep_dir is the directory where the script puts the created files inside
• ref_genome is path to the reference genome fasta file (e.g. hg19.fa)
• dbsnp_file is path to a list of commons snps for the reference genome, such as snp142Common.txt, obtainable from http://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/snp142Common.txt.gz

Prediction

The script predict_region.sh in the directory prediction can be used to predict the copy number of a region.

Usage:

./predict_region.sh [sample_dir] [ref_dir] [prep_dir] [chr] [begin] [end] [ks_threshold] [neighbours_threshold] [fraction] 

• sample_dir is the the directory containing the .bam and the .bai files for the sample cfDNA data
• ref_dir is the the directory containing the .bam and the .bai files for the reference cfDNA data
• prep_dir is the the directory containing the files created in the pre-process
• chr is the chromosome of the target region
• begin is the beginning position of the target region
• end is the end position of the target region
• ks_threshold is the threshold for control similarity
• neighbours_threshold is the threshold for the required minimum number of neighbours
• fetal_fraction is the fraction of the cfDNA in the sample that is fetal origin

Example:

./predict_region.sh ~/sample_dir ~/ref_dir ~/prep_dir chr3 10000000 11000000 0.002 5 0.13

Reference

Arbabi, Aryan, Ladislav Rampášek, and Michael Brudno. "Cell-free DNA fragment-size distribution analysis for non-invasive prenatal CNV prediction." Bioinformatics 32.11 (2016): 1662-1669.