RNA STAR memory error

, , ,
1

I got this while using RNA STAR:

EXITING because of fatal ERROR: not enough memory for BAM sorting: 
SOLUTION: re-run STAR with at least --limitBAMsortRAM 6318500801
Jan 13 01:41:07 ...... FATAL ERROR, exiting

Is there any way around this or is it that the file is too large?

2

Dear @biostaring,
You can try to sort the file beforehand with “Samtools sort order of storing aligned sequences”. MAybe that helps. How big is your file?

Kind regards,
Florian

3

Please also let us know which Galaxy server you are using :slight_smile:

4

The paired end seqs are 14 and 11 GB each.

5

I am using the eu server

6

Hi @biostaring,
did you try to sort the BAM files before using STAR as @Flow suggested?

Regards

7

Hi @biostaring

Did you get help for this already and solve the problem?

If not, this is one thing to check if I am understanding the problem correctly: An RNA STAR job is failing during the last step in the processing (sorting the output BAM result).

That can happen when there are a large number of target sequences (chromosomes, scaffolds, transcripts). Up to a hundred or so is fine, around a thousand can sometimes be Ok, but several hundred thousand or more (millions) won’t work. Why? The Samtools sorting job will run out of memory.

If that is your use-case, the solution is to reduce the number of target sequences in the “custom genome” fasta dataset or to use a different tool:

  • If the fasta is some transcript assembly that you created, check that you are removing very short sequences, especially those that would not be able to capture a hit anyway. The minimum length of the target sequences should be at around twice the length of the query sequences – and maybe a bit longer (~ 200 bases) – but you can generate the read lengths, filter the fasta a few different ways, and test to see what works.
  • If you are using a sequencing read dataset as a custom genome (ran fastq > fasta or something similar), that will never work with RNA-Star as far as I know and for a few different reasons both technical and scientific. Instead, use a tool like BLAST, output a tabular type of result (not a BAM), and consider making the parameters very strict or the output can grow very very large for no practical purpose and potentially fail the job.

Just a guess :slight_smile: and hopefully you’ve already resolved the error!