Accessing reads: AlignmentContext and ReadBackedPileup

Geraldine_VdAuweraGeraldine_VdAuwera Posts: 6,467Administrator, GATK Developer admin
edited October 2012 in Developer Zone

1. Introduction

The AlignmentContext and ReadBackedPileup work together to provide the read data associated with a given locus. This section details the tools the GATK provides for working with collections of aligned reads.

2. What are read backed pileups?

Read backed pileups are objects that contain all of the reads and their offsets that "pile up" at a locus on the genome. They are the basic input data for the GATK LocusWalkers, and underlie most of the locus-based analysis tools like the recalibrator and SNP caller. Unfortunately, there are many ways to view this data, and version one grew unwieldy trying to support all of these approaches. Version two of the ReadBackedPileup presents a consistent and clean interface for working pileup data, as well as supporting the iterable() interface to enable the convenient for ( PileupElement p : pileup ) for-each loop support.

3. How do I get a ReadBackedPileup and/or how do I create one?

The best way is simply to grab the pileup (the underlying representation of the locus data) from your AlignmentContext object in map:

public Integer map(RefMetaDataTracker tracker, ReferenceContext ref, AlignmentContext context)
    ReadBackedPileup pileup = context.getPileup();

This aligns your calculations with the GATK core infrastructure, and avoids any unnecessary data copying from the engine to your walker.

If you are trying to create your own, the best constructor is:

public ReadBackedPileup(GenomeLoc loc, ArrayList<PileupElement> pileup )

requiring only a list, in order of read / offset in the pileup, of PileupElements.

From List and List

If you happen to have lists of SAMRecords and integer offsets into them you can construct a ReadBackedPileup this way:

public ReadBackedPileup(GenomeLoc loc, List<SAMRecord> reads, List<Integer> offsets )

4. What's the best way to use them?

Best way if you just need reads, bases and quals

for ( PileupElement p : pileup ) {
  System.out.printf("%c %c %d%n", p.getBase(), p.getSecondBase(), p.getQual());
  // you can get the read itself too using p.getRead()
}

This is the most efficient way to get data, and should be used whenever possible.

I just want a vector of bases and quals

You can use:

public byte[] getBases()
public byte[] getSecondaryBases()
public byte[] getQuals()

To get the bases and quals as a byte[] array, which is the underlying base representation in the SAM-JDK.

All I care about are counts of bases

Use the follow function to get counts of A, C, G, T in order:

public int[] getBaseCounts()

Which returns a int[4] vector with counts according to BaseUtils.simpleBaseToBaseIndex for each base.

Can I view just the reads for a given sample, read group, or any other arbitrary filter?

The GATK can very efficiently stratify pileups by sample, and less efficiently stratify by read group, strand, mapping quality, base quality, or any arbitrary filter function. The sample-specific functions can be called as follows:

pileup.getSamples();
pileup.getPileupForSample(String sampleName);

In addition to the rich set of filtering primitives built into the ReadBackedPileup, you can supply your own primitives by implmenting a PileupElementFilter:

public interface PileupElementFilter {
    public boolean allow(final PileupElement pileupElement);
}

and passing it to ReadBackedPileup's generic filter function:

public ReadBackedPileup getFilteredPileup(PileupElementFilter filter);

See the ReadBackedPileup's java documentation for a complete list of built-in filtering primitives.

Historical: StratifiedAlignmentContext

While ReadBackedPileup is the preferred mechanism for aligned reads, some walkers still use the StratifiedAlignmentContext to carve up selections of reads. If you find functions that you require in StratifiedAlignmentContext that seem to have no analog in ReadBackedPileup, please let us know and we'll port the required functions for you.

Post edited by Geraldine_VdAuwera on

Geraldine Van der Auwera, PhD

Sign In or Register to comment.