I am running diffbind and has built up the dba object. When running dba.count(), there are few samples failed to process and encountered the following error message --
[W::bam_hdr_read] bgzf_check_EOF: No error
[W::bam_hdr_read] bgzf_check_EOF: No error
[E::bgzf_read] Read block operation failed with error -1 after zd of zu bytes
Anyone has any advices? Thank you in advance!
This generally indicates that there's an issue with your BAM files, usually that they are truncated and missing a special block at the end of the file that marks, well, the end of the file.
You can check your bam files with picard ValidateSamFile, which will help you pinpoint whether this is the issue or not. I don't actually think this is a DiffBind issue.
That's what I thought as well. So I did re-run the parallel alignments for all the samples, and still 4 of 14 samples failed in diffbind dba.count() with the same error.
I also did picard as you suggested, it seemed all the 14 samples got similar error as below (numbers vary) --
Error Type Count
ERROR:MATE_NOT_FOUND 202699
ERROR:MISSING_READ_GROUP 1
WARNING:RECORD_MISSING_READ_GROUP 9050819
However, 10 samples are okay with dba.count() and only 4 of them failed.
How are you aligning? Is everything single-end?
I realized that those bgzf read were derived from bgzf.c (in htslib). Is there any way to avoid to run bgzf.c or htslib inside of dba.count()?
Those 4 failed bam files were okay to run other program such as peak calling, etc. I wonder whether those errors not due to bam file itself.
This might be a question better asked on the Bioconductor support site. The devs have to answer you there (or the Bioc maintainers will get on their case), and they will likely have a better idea of why this is occurring than any of us.
Thank you. I posted at bioc site and the problem resolved. Here is the link https://support.bioconductor.org/p/126303/#126472
Log in to answer this question.
