MikeAxtell
commented
4 months ago Thanks Michał for the feedback and questions.
About slow alignments: The large number of libraries and the highly repetitive nature of the barley genome are the sources of the slowness. ShortStack's alignments take care to make best guesses at where to place multi-mapped reads (see Johnson et al. G3). But that care comes at the cost of extra time, especially when there are a lot of repeats. You can make it easier in at least two ways:
-
Run multiple --align_only jobs with just small numbers of libraries at a time. Then combine all of the resulting BAM files.
-
Run in --mmap -r mode ... just guess at where to drop multi-mapped reads. Will be much faster, at the cost of accuracy.
-
Align with another tool (like sensible bowtie defaults) and then merge the BAMs before running ShortStack.
About the slow performance with a single BAM file: With 128 libraries and a giant, repetitive genome, I'm not surprised that it is slow. You may want to call cluster using ShortStack with just a subset of the libraries, instead of all 128.
If your input BAM file has read groups specified, yes, you should receive a Counts.txt file.
Hope this helps.
From: Michał Stanoch @.> Date: Tuesday, April 23, 2024 at 5:47 AM To: MikeAxtell/ShortStack @.> Cc: Subscribed @.***> Subject: [MikeAxtell/ShortStack] Counts.txt file not created when using .bam files as input (Issue #150)
I am trying to use ShortStack for analyzing a large sRNA-seq experiment (132 samples). The alignment part takes a very long time (almost a full week) on a supercomputer server I am working on, despite specifying 12 threads, 20 GB RAM per thread.
The organism is barley (Hordeum vulgare). I use the toplevel genome without any masking and added mitochondrion, nonchromosomal and chloroplast sequences, which makes the reference rather large, but I got poor results when using a masked genome.
Since I can only submit a job for max a week on the server, I only managed to get the .bam files. I thought I could use them for further analysis by using the --bamfiles rather than --readfiles option to get the results and it works fine, but I don't get the Counts.txt file I would need for DE Seq analysis.
Is this an expected limitation when using .bam files as input?
— Reply to this email directly, view it on GitHubhttps://github.com/MikeAxtell/ShortStack/issues/150, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABUJPCK4Z6JNL5HFD3KVIGLY6YU2BAVCNFSM6AAAAABGURAHHKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGI2TQNBSHA3TGNQ. You are receiving this because you are subscribed to this thread.Message ID: @.***>
mikrzol
dirkjanvw
I am trying to use ShortStack for analyzing a large sRNA-seq experiment (132 samples). The alignment part takes a very long time (almost a full week) on a supercomputer server I am working on, despite specifying 12 threads, 20 GB RAM per thread.
The organism is barley (Hordeum vulgare). I use the toplevel genome without any masking and added mitochondrion, nonchromosomal and chloroplast sequences, which makes the reference rather large, but I got poor results when using a masked genome.
Since I can only submit a job for max a week on the server, I only managed to get the .bam files. I thought I could use them for further analysis by using the
--bamfilesrather than--readfilesoption to get the results and it works fine, but I don't get the Counts.txt file I would need for DE Seq analysis.Is this an expected limitation when using .bam files as input?