The current GATK version is 3.6-0

#### Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

# GATK uses wrong bam index when two plausible indices exist

Member Posts: 6
edited September 2013

I get the following error when running the unified genotyper on the attached bam file:

##### ERROR MESSAGE: SAM/BAM file ./MD_CHW_AAC_2699.recalibrated.bam is malformed: Premature EOF; BinaryCodec in readmode; file: /data/itch/winni/proj/marchini/converge/variantCalling/./MD_CHW_AAC_2699.recalibrated.bam

I was using this code:


/usr/local/bin/java -Xmx128G -jar ~/src/GenomeAnalysisTK-2.7-2-g6bda569/GenomeAnalysisTK.jar -T UnifiedGenotyper -I ./MD_CHW_AAC_2699.recalibrated.bam -R /data/1kg/reference_v37d5/hs37d5.fa -o ./test.vcf



There are two bam index files located in the same directory:


MD_CHW_AAC_2699.recalibrated.bam.bai
MD_CHW_AAC_2699.recalibrated.bai



The second index (.bai) is outdated and older than the bam file. The first index (.bam.bai) is newer than the bam file. The first index is valid and the second index is not valid. What I would expect is for GATK to do one of the following:

1. Use the newer of the two index files and run through
2. Use the .bai file and die with an error saying that the index file is older than the bam file

Instead, what I think GATK is doing is checking that one of the two index files is newer than the bam file, and then using the .bai file regardless of which index was in fact the newer of the two.

I see this error in GATK 2.6-5 and 2.7-2. Is this a bug?

Regards,
Warren Kretzschmar

Post edited by wkretzsch on
Tagged:

Hi Warren,

Actually, the GATK first looks for the index that fits the naming convention where the base name is identical, but the extension is 'bai' instead of 'bam'. The file date doesn't matter at that point. So GATK is using the correct index file in your directory according to its name-based convention. Now, if the problem is that the .bai file is older and out of date, you should see a warning in the console output to that effect. That being said, the error suggests that it is your bam file that is damaged.

In any case, the simplest way to diagnose and/or fix this issue is to delete both index files and regenerate a healthy index using Picard or samtools; if the bam is at fault, they will error out too.

Geraldine Van der Auwera, PhD

Hi Warren,

Actually, the GATK first looks for the index that fits the naming convention where the base name is identical, but the extension is 'bai' instead of 'bam'. The file date doesn't matter at that point. So GATK is using the correct index file in your directory according to its name-based convention. Now, if the problem is that the .bai file is older and out of date, you should see a warning in the console output to that effect. That being said, the error suggests that it is your bam file that is damaged.

In any case, the simplest way to diagnose and/or fix this issue is to delete both index files and regenerate a healthy index using Picard or samtools; if the bam is at fault, they will error out too.

Geraldine Van der Auwera, PhD

• Member Posts: 6

Thanks. I can see the warning about the index being older than the bam file now that I look for it. When I delete the outdated index everything works fine.