From Genome Analysis Wiki
Revision as of 16:33, 15 February 2012 by Zhanxw (talk | contribs)
Jump to: navigation, search


The qplot program calculates various summary statistics some of which are plotted in a PDF file. These statistics can be used to assess the sequencing quality of sequence reads mapped to the reference genome. The main statistics are empirical Phred scores which are calculated based on the background mismatch rate. Background mismatch rate is the rate that sequenced bases are different from the reference genome, EXCLUDING dbSNP positions. Other statistics include GC biases, insert size distribution, depth distribution, genome coverage, empirical Q20 count, and so on.

In the following sections, we will guide you through: how to obtain qplot, how to use qplot, example outputs, interactive diagnostic plots, and real applications in which qplot has helped identify sequencing problems.

Where to Find It

You can obtain qplot in two ways:

(1) Download and extract pre-compiled binary as described in Binary Download.

(2) Download source code of qplot and compile it on your own machine. Please follow the instruction in Source Code Distribution on fetching source code and building instructions.

We recommend the first method since the pre-compiled binary should work out of the box.

Binary Download

We have prepared a pre-compiled (under Ubuntu) qplot. You can download it from: qplot.20120213.tar.gz (File Size: 1.7G)

The executable file is under qplot/bin/qplot.

In addition, we provided the necessary input files under qplot/data/ (NCBI human genome build v37, dbSNP 130, and pre-computed GC file with windows size 100).

You can also find an example BAM input file under qplot/example/chrom20.9M.10M.bam. It is taken from the 1000 Genome Project with sequencing reads aligned to chromosome 20 positions 8M to 9M.

Source Code Distribution

We provide source code in [[ ]]. Optionally, you can download example file or data file:

example: example input file, and expected outputs if you following the direction.

resources data: necessary input files for qplot, including NCBI human genome build v37, dbSNP 130, and pre-computed GC file with windows size 100.

You can put above file(s) in the same folder and follow these steps:

  • 1. Unarchive downloaded file
tar zvxf qplot-source.20120213.tar.gz

A new folder qplot will be created.

  • 2. Build libStatGen
cd qplot
make libStatGen

This step will download a necessary software library [libStatGen] and compile source code into a binary code library.

  • 3. Build qplot
make all

This step will then build qplot. Upon success, the executable qplot can be found under qplot/bin/.

  • 4. (Optional) unarchive example data
tar zvxf qplot-example-data.tar.gz

Example data will be extracted to qplot/data/.


Command line

After you obtain the qplot executable (either by compiling the source code or by downloading the pre-compiled binary file), you will find the executable file under qplot/bin/qplot.

Here is the qplot help page:

 some_linux_host > qplot/bin/qplot

 The following parameters are available.  Ones with "[]" are in effect:

             References : --reference [/data/local/ref/karma.ref/human.g1k.v37.umfa],
                          --dbsnp [/home/bingshan/data/db/dbSNP/dbSNP130.UCSC.coordinates.tbl],
                          --gccontent [/home/bingshan/data/db/gcContent/gcContent.hg37.w250.out]
  Create gcContent file : --create_gc [], --winsize [100]
            Region list : --regions [], --invertRegion
           Flag filters : --read1_skip, --read2_skip, --paired_skip,
         Dup and QCFail : --dup_keep, --qcfail_keep
        Mapping filters : --minMapQuality [0.00]
     Records to process : --first_n_record [-1]
       Lanes to process : --lanes []
           Output files : --plot [], --stats [], --Rcode []
            Plot labels : --label [], --bamLabel []

Input files

qplot runs on the input BAM/SAM file(s) specified on the command-line after all other parameters.

Additoinally, three (3) precomputed files are required.

  • --reference

The reference genome is the same as karma reference genome. If the index files do not exist, qplot will create the index files using the input reference fasta file.

  • --dbsnp

This file has two columns. First column is the chromosome name which must be consistent with the reference created above.

  • --gccontent

Although GC content can be calculated on the fly each time, it is much more efficient to load a precomputed GC content from a file. To generate the file, use the following command:

qplot --rerefence reference.fa --windowsize winsize --create_gc reference.gc

Note: Before running qplot, it is critical to check how the chromosome numbers are coded. Some bam files use just numbers, others use chr + numbers. You need to make sure that the chromosome numbers from the reference and dbSNP are consistent with the bam file.


Some of the command line parameters are described here, but most are self explanatory.

  • Flag filter

By default all reads are processed. If it is desired to check only the first read of a pair, use --read2_skip to ignore the second read. And so on.

  • Duplication and QCFail

By default reads marked as duplication and QCFail are ignored but can be retained by



  • Records to process

The --first_n_record option followed by a number, n, will enable qplot to read the first n reads to test the bam files and verify it works.

  • Lanes to process

If the input bam files have more than one lane and only some of them need to be checked, use something like --lanes 1,3,5 to specify that only lanes 1, 3, and 5 need to be checked.

NOTE In order for this to work, the lane info has to be encoded in the read name such that lane number is the second field with the delimiter ":".

  • Mapping filters

Qplot will exclude reads with lower mapping qualities than the user specified parameter, --minMapQuality. By default, all reads will be included in analysis.

  • Region list

If the interest of qplot is a list of regions, e.g. exons, this can be achieved by providing a list of regions. The regions should be in the form of "chr start end label" each line in the file (NOTE: start and end position are inclusive and they follow the convention of BED file). In order for this option to work, within each chromosome (contig) the regions have to be sorted by starting position, and also the input bam files have to be sorted. For example, you can create a text file, region.txt like following:

1 100 500 region_A
1 600 800 region_B
2 100 300 region_C

Then specifying --regions region.txt enables qplot calculate various statistics out of sequenced bases only from the above 3 regions.

Qplot also provides the --invertRegion option. Enabling this option tells qplot to operate on those sequence bases that are outside the given region.

  • Plot labels

Two kinds of labels are enabled. --label is the label for the plot (default is empty) which is prepended to the title of each subplot. --bamLabels followed by a column separated list of labels provides the labels for each input SAM/BAM file, e.g. sample ID (default is numbers 1, 2, ... until the number of input bam files). For example:

--label Run100 --bamLabels s1,s2,s3,s4,s5,s6,s7,s8
  • Multiple threading (not officially supported)

Number of concurrent threads running for the input bam files. One bam file at a time will be processed by one thread. Therefore using a number which is dividable by the number of input bam files will make it more efficient. One extra thread requires about 375Mb more memory on top of the around 4Gb of memory used to hold reference and GC content files.

Output files

There are three (optional) output files.

  • --plot qa.pdf

Qplot will generate a PDF file named qa.pdf containing 2 pages each with 4 figures. If --pages 1 is specified, only page 1 is output. The plot is generated using Rscript.

  • --stats qa.stats

Qplot will generate a text file named qa.stats containing various summary statistics for each input BAM/SAM file.

  • --Rcode qa.R

Qplot will generate qa.R which is R code used for plotting the figures in qa.pdf file. If Rscript is not installed in the system, you can use the qa.R to generate the figures on other machines, or extract plotting data from each run and combine multiple runs together to generate more comprehensive plots (See Example).


Qplot can generate diagnostic graphs, related R code, and summary statistics for each SAM/BAM file.

Built-in example

In the pre-compiled binary download, you will find a subdirectory named examples. We provide a sample file from the 1000 Genome project, it contains aligned reads on chromosome 20 from position 8 Mbp to 9Mbp. You can use qplot using the following commandline:

../bin/qplot --reference ../data/human.g1k.v37.umfa --dbsnp ../data/dbSNP130.UCSC.coordinates.tbl --gccontent ../data/human.g1k.w100.gc --plot qplot.pdf --stats qplot.stats --Rcode qplot.R --label "chr20:9M-10M" chrom20.9M.10M.bam

Sample outputs are listed below:

Figure: qplot.pdf

Summary statistics:

Stats\BAM       chrom20.9M.10M.bam
TotalReads(e6)  1.11
MappingRate(%)  97.24
MapRate_MQpass(%)       97.24
TargetMapping(%)        0.00
ZeroMapQual(%)  2.39
MapQual<10(%)   2.86
PairedReads(%)  83.76
ProperPaired(%) 71.34
MappedBases(e9) 0.04
Q20Bases(e9)    0.04
Q20BasesPct(%)  88.63
MeanDepth       42.22
GenomeCover(%)  0.03
EPS_MSE 1.81
EPS_Cycle_Mean  18.71
GCBiasMSE       0.01
ISize_mode      137
ISize_medium    184
DupRate(%)      5.90
QCFailRate(%)   0.00
BaseComp_A(%)   29.9
BaseComp_C(%)   20.1
BaseComp_G(%)   20.2
BaseComp_T(%)   29.8
BaseComp_O(%)   0.1

Gallery of examples

Here we show qplot can be applied in various sequencing scenarios. Also users can customize statistics generated by qplot to their needs.

  • Whole genome sequencing with more than one lane



Summary statistics text file

TotalReads(e6)  72.94   64.52   74.87   62.25   67.21
MappingRate(%)  97.62   97.51   97.75   97.52   97.35
MapRate_MQpass(%)       97.62   97.51   97.75   97.52   97.35
TargetMapping(%)        45.53   45.51   46.39   45.81   46.23
ZeroMapQual(%)  11.52   11.64   11.77   10.97   11.14 
MapQual<10(%)   11.91   12.03   12.17   11.34   11.52
PairedReads(%)  100.00  100.00  100.00  100.00  100.00
ProperPaired(%) 96.14   96.10   96.34   95.60   95.91 
MappedBases(e9) 2.11    1.87    2.20    1.82    1.97
Q20Bases(e9)    2.05    1.81    2.13    1.76    1.91
Q20BasesPct(%)  97.12   96.98   96.75   96.90   96.91
MeanDepth       35.08   31.05   36.55   30.30   32.82
GenomeCover(%)  2.10    2.10    2.10    2.09    2.10
EPS_MSE 8.89    6.88    8.50    13.32   6.86
EPS_Cycle_Mean  26.04   25.88   25.86   26.12   25.77
GCBiasMSE       0.04    0.05    0.04    0.07    0.04
ISize_mode      250     250     249     210     250
ISize_medium    271     270     270     260     270 
DupRate(%)      3.50    3.79    3.27    4.51    3.56
QCFailRate(%)   0.00    0.00    0.00    0.00    0.00
BaseComp_A(%)   26.3    26.4    26.3    26.8    26.3
BaseComp_C(%)   23.7    23.6    23.7    23.2    23.7
BaseComp_G(%)   23.2    23.0    23.2    22.7    23.1
BaseComp_T(%)   26.8    27.1    26.8    27.3    26.9
BaseComp_O(%)   0.0     0.0     0.0     0.0     0.0

  • Whole genome sequencing with 24-multiplexing

With a customized script, we aggregated 24 bar-coded samples in the same graph. The graph will help compare sequencing quality between samples.

QPlot of 24 samples(PDF)

  • Interactive qplot

Qplot can be interactive. In the following example, you can use mouse scroll to zoom in and zoom out on each graph and pan to a certain part of the graph. By presenting qplot data on a web page, users can easily identify problematic sequencing samples. Users of qplot can customize its outputs into webpage format greatly easing the data exploring process.

QPlot of 24 samples(HTML)

Diagnose sequencing quality

Qplot is designed and implemented for the need of checking sequencing quality. Besides the example of analyzing RNA-seq data as shown in our manuscript, here we demonstrate two additional scenarios in which qplot can help identify problems after obtaining sequencing data.

  • Base quality distributed abnormally

Example of qplot helping to identify wrong phred base quality

By checking the first graph "Empirical vs reported Phred score", we found reported base qualities are shifted to the right. In this particular example, '33' was incorrectly added to all base qualities. When such data used in variant calling, we may increase false positive SNP variants.

  • Bar-coded samples

Example of qplot identifying the effect of ignoring bar-coding

By checking "Empirical phred score by cycle" (top right graph on the first page), we noticed the empirical qualities in the first several cycles are abnormally low. This phenomenon leads us to hypothesize that the first several bases have different properties. Further investigation confirmed that this sequencing was done using bar-coded DNA samples, but the analysis did not properly de-multiplex each sample.


Questions and requests should be sent to Bingshan Li (bingshan@umich.edu)