VAAL Manual

VAAL manual (version 1.1, 10/23/08)

Overview

VAAL is a polymorphism discovery algorithm for short reads. To run it, you provide reads (and quality scores) from a ‘sample genome’ as input, along with a vector sequence to trim from the reads, and a reference sequence for a related genome to compare to. VAAL produces as output a an assembly for the sample genome, together with a mask showing which bases are ‘trusted’. VAAL then deduces from that a list of differences between the sample and related genomes.
Alternatively, VAAL can be provided as input read data for two sample genomes, together with a reference sequence for a related genome. In this case, VAAL produces assemblies for each of the sample genomes, and compares them to each other, thereby deducing a list of differences between them.
VAAL has been tested on bacteria, using single lanes of 36 bp unpaired reads from the Illumina platform.
 

Software availability

Download VAAL from:

ftp://ftp.broadinstitute.org/pub/crd/VAAL/latest_source_code/LATEST_VERSION.tar.gz

 

System requirements


For bacterial genomes, VAAL uses roughly 10 GB of memory.

VAAL should work on most x86-64 Linux systems, but some changes may be needed.

We require that GCC version 4.3.0 or later be installed on your system.

Building VAAL

See the general build instructions here:

http://www.broadinstitute.org/science/programs/genome-biology/computational-rd/general-instructions-building-our-software

In a nutshell, you'll untar, configure, make, and make install:

tar xvzf LATEST_VERSION.tar.gz
cd vaal-nnnnn
./configure -prefix=<your-install-directory>
make
make install

If all goes well, you'll find the VAAL executables in <your-install-directory>/bin.
Add that directory to your path, and you're ready to run VAAL.
 

Preparing the inputs

• Pick a directory dir where you will perform the calculations.
• Pick a project name. We’ll call this reads, but you may substitute any string that does not have a dot or slashes in it.
• Inside dir, create a directory named reads.
• Inside dir/reads, create two files reads.fasta and reads.qual. These may be links. The reads.fasta file should be in fasta format. The reads.qual file should be formatted like a fasta file, but instead of bases, have integer quality scores (between 0 and 255), separated by spaces.
• Inside dir/reads, create a file ref.fasta, the reference sequence for the related genome. You may use anything in place of ref (but no slashes, please). This may also be a link.
 

Running VAAL

Change directory to dir, and type VAALrun SOURCE=reads REF=ref VECSEQ=v where reads and ref are as above, without fasta extensions, and v is a vector sequence to be trimmed from the right side of the reads. For example, an appropriate value for unpaired Illumina reads is at present AGATCGGAAGAGCTCGTATGCCGTCTTCTGCTTGAAAAAAAAAAAAAAAAAAAAAAAAA.
 

Optional arguments.

• You can see all arguments and default values by typing simply ‘VAALrun’.
• If you specify RUN=runname, runname will be the name of the run. Otherwise, your user name is used as the name of the run. See below for how the name of the run is used.
• If you specify TRUNC=n, reads will be truncated to n bases.
• If you specify TRUE_REF=sample, VAAL will look for a file sample.fasta in dir/reads, and use it to assess the correctness of its results. Of course to do this you need to be carrying out a controlled experiment for which you know exactly what the genome of the sample is in advance.

Finding the results directory. In dir/reads, VAAL will create a subdirectory vs_ref, where ref is the name you used for the related reference sequence. Inside this directory is a directory runname (see optional argument RUN, above). This directory contains the results. If you specified the TRUNC argument, the vs_ subdirectory will instead be inside another subdirectory trunc_n.

Finding the assisted assembly. In the results directory, there will be files reads.assembly.fasta and reads.assembly.trusted.txt. The first is the assembly of the sample genome in fasta format, and the second is a human-readable fasta-style ‘vecbitvector’ showing which bases on the assembly are ‘trusted’.

Finding the differences. Inside the results directory is a file reads.VAAL4.K28.out, containing the list of all observed differences (SNPs or indels) between the sample genome and the related genome. Composite events are decomposed into their constituent SNPs and indels. There are two header lines, followed by one line per difference, followed by summary statistics. A difference line for a SNP looks like

0 20895 left=TGGATCAGACCTTTAGCAGC sample=C ref=G right=TGACGGTCCACGATCGGTTG

which is interpreted as follows: on zero-based reference contig 0, at zero-based position 20895, the sample shows C but the reference shows G. The ‘left’ and ‘right’ arguments specify 20 flanking bases on each side, which may be used as a sanity check on coordinates. For a two-base insertion in the sample, one might see instead

0 2352451 left=GACCTGGCGAATCGCCTCTT sample=CG ref= right=TTTGTCTCCGGCCTTTTCGC

indicating that CG is inserted at the given position. For deletions, ‘sample’ is empty but ‘ref’ is not.

Alternatively, the file reads.VAAL4.K28.compound.assembly exhibits all polymorphisms, sorted into groups, with composite events grouped together.

Finding bases that are not different: Inside the results directory is a file reads.VAAL4.K28.nonpolymorphic.bitvector. This is a binary file with one bit per base on the reference, stored as a ‘vecbitvector’ object (defined in the codebase). A bit is ‘on’ if the corresponding base in the sample appears to agree with the corresponding base in the reference. Note that if there is an insertion in the sample, the surrounding bits in the reference may still be on. In addition, for compound events, bases internal to the compound event will not be recording as agreeing with the reference.

Comparison to truth data: If you specified the TRUE_REF argument, there will be a file reads.VAAL4.K28.compound.truepoly, providing a list of all ‘callable’ polymorphisms, in the same format as the compound.assembly file, a file reads.VAAL4.K28.compound.events, listing all missed compound events and false positive calls, and a file reads.VAAL4.K28.compound.subevents containing some summary statistics.

Comparing two assemblies: If you have sequenced two samples, for which you have a common reference sequence, you can compare them using
TruePoly REFA=reads1.assembly.fastb TRUSTEDA=reads1.assembly.trusted
REFB=reads2.assembly.fastb TRUSTEDB=reads2.assembly.trusted
where the reads... arguments are full paths of files in the results directories for the two samples. The coordinates of the polymorphisms will be on the sample 2 assembly. Composite events are decomposed into their constituent SNPs and indels.