CoverageCompacter

Next generation sequencing (NGS) is an expensive process but its cost could be significantly reduced if only good quality libraries were sequenced. CoverageCompacter can be used to generate estimates of the coverage of NGS libraries that have been sequenced at ultra low depth if those libraries were to undergo further sequencing. This is useful for analysing DNA which is likely to have regions of allelic dropout such as samples that have undergone whole genome amplification (WGA) from single celled sequencing experiments or other experiments that use low amounts of starting material such as DNA extracted from laser capture microdissection experiments. The output from CoverageCompacter can be used to guide which libraries to shortlist for further sequencing or producing useful summaries of sequencing depth in the BED format.

Principles and applications

This software was inspired by work by Zhang et al 2015 (see link to original article below). NGS libraries produced from DNA that has undergone WGA are likely to contain amplificaton bias, i.e. areas of the genome which are over/under represented. This bias can be estimated from low pass sequencing data based on 2 principles as discussed by Zhang et al. (1) Essentially sequencing a library at ultra low sequencing can be thought of as 'sampling' a library, we are sequencing regions which are well represented in the overall fragment pool. (2) WGA tends to produce amplicons rangeing between 10-50Kb in size. Infering from these 2 concepts; if a library is sequenced at low depth and 1 or more reads map to a given loci, then it can be asumed that if more sequencing is done then its likely to cover the 10-50kb region surrounding the loci.

• Original aritcle: https://www.nature.com/articles/ncomms7822
• Pubmed link: https://www.ncbi.nlm.nih.gov/pubmed/25879913

CoverageCompacter uses the BED format output of Samtools depth (see Samtools docs for detals) and compacts this output into smaller loci. CoverageCompacter takes the binsize as an argument and will compact areas of coverage into a single loci surounded by areas of no coverage up to half the size of the binSize arguement. The reasoning behind this is that up to half an amplicon telomeric and centromeric of a given read are likely be contained within the library but not yet sequenced. Where 2 amplicons are separated by less than half a bin, CoverageCompacter will merge the loci into a single loci. See 'Inputs and outputs of the software' examples 1-4 for more detail about how this is implemented. The output files can then be more easily compared with established tools such as 'bedtools intersect'.

CoverageCompacter can also be used to compress loci with a minimum depth chosen by the user. This can be suplied as an arguement. The 'NoCov' arguement is set to 0 by default but if set to 10 (for example) then loci will be identified as having sufficient coverage if they have greater than 10 reads covering a single base. The loci will be compressed into regions with above or below the minimum coverage and separated by the binsize. This is useful for identifying regions that may require more sequencing, or comparing multiple libraries to identify areas of poor coverage across a range of samples. It can also be used to identify regions that have excessive coverage or verifying that targetted sequencing experiments have performed appropriately. It can also be used to identify transcript boundaries in RNASeq datasets.

Dependencies

Analysis ready BAM files Reference genome Samtools (we used samtools v1.3.1) Python 3

Installation

git clone https://github.com/wes3985/CoverageCompacter.git

Instructions for usage

(1) Run samtools depth over your BAM file(s) to generate a depth file. You can use the following command and launch each chromosome to a separate core on a cluster.

bam_file_list = a text file containing a list with the full paths to your bam files. path_to_ref = the full path to the reference genome. chr = the chromosome in the same format as your reference. depth_file_full_outpath = the full file path to write the depth data to.

samtools depth -a -q 10 -Q 20 --reference $path_to_ref -r $chr -f $bam_file_list > $depth_file_full_outpath

(2) Call CoverageCompacter, this can take some time for large genomes and/or many samples, we recommend launching each chromosome to a different core.

in_depthfile = the input depth file produced by samtools depth outfile = the full file path to write the depth data to. samples = a string containing the samples separated by commas, if calling from a python shell: "sample1,sample2,sample3" CHROM = The chromosome covered by the infile e.g "chr1", "1". The infile must span only a single chromosome or loci from sequential chromosomes will be merged. This functionality may be improved in future versions depending on demand. binSize = the presumed size of the amplicons, recommend size is 10000 based on Zhang et al (see link above). Setting the binSize to zero also makes this a useful tool to check for transcript boundaries in RNASeq data. NoCov = the minimum depth at which will be regions will be separated into coverage / no coverage

from a UNIX command line or launch script

python CoverageCompacter $in_depthfile $outfile $samples $CHROM $binSize

from a python interface

from CoverageCompacter import CoverageCompacter CoverageCompacter(depth_file, outfile, samples, CHROM, 10000, 0)

Inputs and outputs of the software

INPUT EXAMPLE: This is the output format of samtools depth, col1 = chromosome, col2 = position in chromosome, col3 = depth at position

See samtools depth docs for more information

chr1 1 0 chr1 2 0 chr1 3 0 chr1 4 0 chr1 5 0 chr1 6 0 chr1 7 0 chr1 8 0 chr1 9 1 chr1 10 1 chr1 11 1 chr1 12 1 chr1 13 1 chr1 14 1 chr1 15 1

OUTPUT EXAMPLE

chr start end size firstCoveredBase lastCoveredBase meanCoverage NBasesCovered DepthSum coverageFraction chr1 1 8 8 None None 0 0 0 0 chr1 9 15 7 9 15 1 7 7 1

The output file is a headered file in the bed format. Below is a description of each header:

chr = chromosome start = start position of the loci end = end postion of the loci size = total size of the loci firstCoveredBase = the first base in the loci with coverage >0 lastCoveredBase = the last base in the loci with coverage >0 meanCoverage = mean coverage accross the loci NBasesCovered = toal number of bases covered in the loci DepthSum = the total sum of depth in the loci, this is useful for identifying regions of high relative coverage coverageFraction = fraction of the loci that has coverage >0

INPUT EXAMPLE 1

chr1 1 0 chr1 2 0 chr1 3 1 chr1 4 1 chr1 5 1 chr1 6 0 chr1 7 0 chr1 8 0 chr1 9 1 chr1 10 1 chr1 11 0 chr1 12 0 chr1 13 1 chr1 14 1 chr1 15 0 chr1 16 1 chr1 17 1 chr1 18 1 chr1 19 0

(1) OUTPUT of CoverageCompacter with binSize set to '0'

Notes:

- areas of no coverage are joined into a single loci in the bed format
- areas of continuous coverage are also joined into single loci

chr start end size firstCoveredBase lastCoveredBase meanCoverage NBasesCovered DepthSum coverageFraction chr1 1 2 2 None None 0.0 0 0 0.0 chr1 3 5 3 3 5 1.0 3 3 1.0 chr1 6 8 3 None None 0.0 0 0 0.0 chr1 9 10 2 9 10 1.0 2 2 1.0 chr1 11 12 2 None None 0.0 0 0 0.0 chr1 13 14 2 13 14 1.0 2 2 1.0 chr1 15 15 1 None None 0.0 0 0 0.0 chr1 16 18 3 16 18 1.0 3 3 1.0 chr1 19 19 1 None None 0.0 0 0 0.0

(2) OUTPUT of CoverageCompacter with binSize set to '1'

Notes:

- The boundary of the bin is moved to binSize/2 and rounded up to the nearest whole number - '1' in this case which means that loci
with coverage will include 1 bp with zero coverage on either side. Where a single base has no coverage between 2 loci where coverage
is present these adjacent loci will be merged.

chr start end size firstCoveredBase lastCoveredBase meanCoverage NBasesCovered DepthSum coverageFraction chr1 1 1 1 None None 0.0 0 0 0.0 chr1 2 6 5 3 5 0.6 3 3 0.6 chr1 7 7 1 None None 0.0 0 0 0.0 chr1 8 11 4 9 10 0.5 2 2 0.5 chr1 12 19 8 13 18 0.625 5 5 0.625

(3) OUTPUT of CoverageCompacter with binSize set to '2'

Notes:

- The output is the same as supplying binsize of 1 (2/2=1) because a boundary is formed at half the bin size and a single base cannot
be cut in half. 

chr start end size firstCoveredBase lastCoveredBase meanCoverage NBasesCovered DepthSum coverageFraction chr1 1 1 1 None None 0.0 0 0 0.0 chr1 2 6 5 3 5 0.6 3 3 0.6 chr1 7 7 1 None None 0.0 0 0 0.0 chr1 8 11 4 9 10 0.5 2 2 0.5 chr1 12 19 8 13 18 0.625 5 5 0.625

(4) OUTPUT of CoverageCompacter with binSize set to '3'

Notes:

• The area between postion 6 and postion 8 has 3 bp with zero coverage, in cases where splits occur that could reasonably fit into both adjacent loci, the loci is split the extra non-covered base is allocated to the previous loci

chr start end size firstCoveredBase lastCoveredBase meanCoverage NBasesCovered DepthSum coverageFraction chr1 1 7 7 3 5 0.42857142857142855 3 3 0.42857142857142855 chr1 8 19 12 9 18 0.5833333333333334 7 7 0.5833333333333334

INPUT EXAMPLE 2:

chr1 1 0 chr1 2 0 chr1 3 1 chr1 4 2 chr1 5 1 chr1 6 0 chr1 7 0 chr1 8 0 chr1 9 1 chr1 10 1 chr1 11 2 chr1 12 2 chr1 13 1 chr1 14 1 chr1 15 0 chr1 16 1 chr1 17 1 chr1 18 1 chr1 19 0

(4) OUTPUT of CoverageCompacter with binSize set to '0' and noCov set to '1'

chr start end size firstCoveredBase lastCoveredBase meanCoverage NBasesCovered DepthSum coverageFraction chr1 1 3 3 None None 0.0 0 0 0.0 chr1 4 4 1 4 4 2.0 1 2 1.0 chr1 5 10 6 None None 0.0 0 0 0.0 chr1 11 12 2 11 12 2.0 2 4 1.0 chr1 13 19 7 None None 0.0 0 0 0.0

Future Updates

(1) Add an argument which can be specified by the user to exclude regions of no coverage from the output file.

Support

Please inform us of any issues at 'w.woollard@ucl.ac.uk'