Size for sjdbOverhang for multiple read length datasets

[santiagorevale]

Hi Alex,

I was wondering which value should I specify when creating the reference for the --sjdbOverhang parameter if I want to use the same reference index against 2x75, 2x101, and 2x151 bp long datasets.

Following the documentation, it says to use max(ReadLength)-1m which in my case was 150, but then when I try to run the alignment for one of the shorter read length datasets, I get the error:

Jun 05 11:28:04 ..... started STAR run
Jun 05 11:28:04 ..... loading genome

EXITING because of fatal PARAMETERS error: present --sjdbOverhang=100 is not equal to the value at the genome generation step =150
SOLUTION: 

Jun 05 11:28:04 ...... FATAL ERROR, exiting

I'm creating the reference in with the following command:

STAR \
    --runMode genomeGenerate \
    --genomeDir STARindex \
    --genomeFastaFiles genome.fa \
    --sjdbGTFfile genes.gtf \
    --sjdbOverhang 150 \
    --runThreadN 24

and I'm running the aligner with the following command:

STAR \
    --genomeDir STARindex \
    --runThreadN 8 \
    --readFilesIn Sample01_S1_L001_R1_001.fastq.gz Sample01_S1_L001_R2_001.fastq.gz  \
    --readFilesCommand "gunzip -c" \
    --sjdbOverhang 100 \
    --outSAMtype BAM SortedByCoordinate \
    --twopassMode Basic \
    --limitBAMsortRAM 112000000000 \
    --limitOutSJcollapsed 1000000 \
    --outFileNamePrefix Sample01.

Software version is 2.7.3a on a CentOS 7.6.1810 OS.

Thank you very much in advance.

Cheers,
Santiago

[alexdobin]

Hi Santiago,

at the mapping step, you do not need to specify --sjdbOverhang 100 , if annotations were already included at the genome generation step.

Cheers
Alex

[santiagorevale]

Hi Alex,

Thanks! So, for the current scenario where I'll be using 75, 100, and 150 pair-end reads, should I format the reference using --sjdbOverhang 150 and it'll work fine for all of them? As if I would have created a specific reference with a --sjdbOverhang of 75, 100, 150, for each of them?

Thanks again!

Cheers,
Santiago

[alexdobin]

Hi Santiago,

one fixed value of --sjdbOverhang will not work the same as a specific reference for each read length. The latter option, strictly speaking, is more accurate, but the actual effect is marginal. Nowadays I would recommend using 74 for the 75,100,150 read lengths samples.

Cheers
Alex

[santiagorevale]

Thanks for the clarification!

Cheers,
Santiago

[zhangpeng1202]

Hi, Alex,

I noticed in your STARmanual, it says: " In case of reads of varying length, the ideal value is max(ReadLength)-1."
If I understand correctly, for the data including 75, 100, and 150 read lengths. The ideal --sjdbOverhang should be set to max(75, 100, 150)-1, which is 149?

Another quick question, I realize sometimes the RNAseq read length could be 76, 101, 151. Which --sjdbOverhang should we use in generating genome index? Should we generate genome index for 75, 76, 100, 101, 150, 151 separately, and using 74, 75, 99, 100, 149, 150 as --sjdbOverhang respectively?

Thanks very much!
Peng

[alexdobin]

Hi Peng,

using the "non-ideal" value of --sjdbOverhang leads only to marginal changes in most cases, so you could simply use the default value of 100.