the --sjdbOverhang parameter in generating index

[Iris Zhu]

Hi, I am new to STAR. I am trying to build hg19 index of STAR. There is a parameter --sjdbOverhang, which is supposed to  Readlength-1. My question is:  Do I need to re-build the index each time for sequence reads of a different length?

For example, this time my reads are 100 nt so I use --sjdbOverhang 99, next time my reads are 75 nt then I need to build a different index with --sjdbOverhang 74?

Thank you for your help,

Iris

[Iris Zhu]

Could anybody answer my question? I would really appreciate it. 

I have multiple RNA-seq data sets to analyze, they are of different read lengths, 65, 75, 100, 125 nt. To build 4 different index is not a big deal but I just want to make sure that is what's needed, if I want to use STAR.

Thank you,

Iris 

[Alexander Dobin]

Hi Iris,

you can simply use the default value (100) for all of your cases. This parameter does make almost no difference for reads longer than 50b.

Cheers

Alex

[Sarah Young]

Hi Alex,

We have the same issue, except that we need a reference we can use on 25bp and lengths up to 50bp and beyond.   What do you recommend in that case?  Should we use multiple index builds?

Thanks for your help!

~Sarah

[Alexander Dobin]

Hi Sarah,

for 25b reads, I would recommend generating a separate index with --sjdbOverhang 24.

For reads longer than 50b, you can use --sjdbOverhang 100

Cheers

Alex

[Jean Chang]

I have an hg19 index built with --sjdbOverhang 24 but recently received 38bp reads. Will I get significantly better alignments if I re-build my index with --sjdbOverhang 37 than if I continue to use my existing index? This thread indicates it should not matter much for reads >50bp but is the impact significant enough for my shorter reads that I should generate new indices for different read lengths if under 50bp?

Thanks,

Jean

[Alexander Dobin]

Hi Jean,

in your case, I would recommend generating the new index. In general, the indexes for longer reads will work fine for shorter reads, but not vice versa.

Cheers

Alex