Copy Number Variation

The BCFtools package implements two methods (the polysomy and cnv commands) for sensitive detection of copy number alterations, aneuploidy and contamination. In contrast to other methods designed for identifying copy number variations in a single sample or in a sample composed of a mixture of normal and tumor cells, this method is tailored for determining differences between two cell lines, which allows to distinguish between normal and novel copy number variation.

Installation

The polysomy command requires the GNU Scientific Library. As described in the installation, it must be compiled with USE_GPL=1. If it is still not visible in the list of commands, recompile with:

make USE_GPL=1 clean all

Now typing bcftools polysomy should give you a list of available options.

Preparing input data

The polysomy command takes on input VCF with FORMAT columns annotated with B-Allele Frequency (the BAF annotation). The cnv command in addition requires the presence of Log R Ratio values (the LRR annotation). If the experimental data were prepared by Illumina’s GenomeStudio, its text output can be converted to VCF using the fcr-to-vcf script. Please check this usage example for details and some test data to experiment with.

Detecting aneuploidy and contamination

Large aberrations which affect whole chromosomes, such as aneuploidy or contamination, can be discerned directly from the overall distribution of BAF values. The command is

bcftools polysomy -v -o outdir/ file.vcf

and the results can be found in outdir/dist.dat. The file can be inspected visually or processed by standard unix commands. For example, a list of chromosomes which are aberrant or uncertain can be obtained by

cat outdir/dist.dat | awk '$1=="CN" && $3!=2.0'

For clean data, the third column should be 2.0 for normal diploid state, 1.0 for a loss, 3.0 for gain, and -1 is used when the program cannot determine the state, usually because of noisy data. If uncertain, it is very useful to inspect the BAF distribution by eye. The distribution can be plotted using the auto-generated matplotlib script

python outdir/dist.py

When the goodness-of-fit threshold -f is set too strict or when the experimental intensities are too different from the expected distribution, the fit may fail. This is indicated by printing -1 instead of a copy number state. If the program outputs a non-diploid state on multiple chromosomes, this may indicate contamination or very noisy input data.

Example of the graphical output from the polysomy command. In this figure, the sample had three copies of chromosome X.

Detecting subchromosomal CNVs

The strength of the CNV caller is in the pairwise calling mode which was designed to detect differences between two samples. This greatly helps to reduce the number of false calls and also allows one to distinguish between normal and novel copy number variation. The command is

bcftools cnv -c control_sample -s query_sample -o outdir/ -p 0 file.vcf

The -p 0 option tells the program to automatically call matplotlib and produce plots like the one in this example:

Example of the graphical output from the cnv command. In this figure, the control sample is normal and the query has two CNVs: a duplication and a loss.

Working with non-Illumina data

If the fcr-to-vcf script fails from some reason or in case the input data is in a different format, the VCF file can be annotated "manually":

# Annotation file with BAF values for two samples
$ zcat baf.txt.gz | head -2
11  193096  0.24    0.16
11  193194  0.61    0.81

# Index the annotation file and fill in the BAF values. For the latter, we need to
# add a BAF definition into the VCF header
$ tabix -s1 -b2 -e2 baf.txt.gz
$ echo '##FORMAT=<ID=BAF,Number=1,Type=Float,Description="NGS estimate of BAF">' > baf.hdr
$ bcftools annotate -a baf.txt.gz -h baf.hdr -c CHROM,POS,FMT/BAF -Ob -o output.bcf input.bcf

References

Please cite our paper when using our software: http://www.ncbi.nlm.nih.gov/pubmed/27176002

Feedback

We welcome your feedback, please help us improve this page by either opening an issue on github or editing it directly and sending a pull request.