Amplicon data

[mcgml]

Hi

Looks like a great tool I just wondered it will work with amplicon data. I seem to recall haplotype caller struggling with the lack of sequence diversity.

Thanks Matt

[mozack]

Yes, ABRA will likely have problems with the lack of sequence diversity also. We do have an update in progress that should help with this. If you are feeling adventurous, we could try get you going on an alpha version of the new software.

[mcgml]

Thanks for your reply, I would be very grateful to try the new functionality. I have a specific case where the variant caller is unable to find a 20bp deletion but it can be seen clearly in igv with soft clipping enabled. How do I get my hands on the new version.

Thanks Matt

[mozack]

An alpha release of ABRA2 is available. See: https://github.com/mozack/abra2/releases Download and extract abra2_0.1.tgz

Sample command line:

JAR=abra2/release/test/abra2_0.1-SNAPSHOT-jar-with-dependencies.jar
# Path to directory containing the .so files
ABRA_LIBDIR=abra2/release/test
REF=ref/hg19.fa
NORMAL=normal.sort.bam
TUMOR=tumor.sort.bam
BED=uncseq5.bed

java -Xmx4G -Djava.library.path=$ABRA_LIBDIR -jar $JAR --ref $REF --in $NORMAL,$TUMOR --out normal.abra.bam,tumor.abra.bam --working abra_temp_dir --targets $BED --threads 8 --sa > abra2.log 2>&1

[mcgml]

Thanks Lisle ABRA worked well. See IGV screenshot below. Top pane is ABRA, bottom is original bwa.

Could you explain how the amplicon function works please? Are you still using a de bruijn graph?

Thanks Matt

[mozack]

The --sa param you specified disables the de Bruijn graph assembly. In ABRA2, we've introduced local Smith Waterman alignment for reads with high base quality soft clipping. Optimal use cases for one method over the other (or both combined) has not yet been determined (although cases such as yours will clearly be better served by Smith Waterman).