Indexing- Exons and splice sites

[New2Bioinfo]

Hi,
I have been following the github tutorial, https://github.com/griffithlab/rnaseq_tutorial/wiki

to learn RNAseq.

I was on the indexing step and it says to first export exons and splice sites from reference genome using in-built python scripts before starting the indexing. And then that this information will be used during alignment.

It would be great if someone could explain the rationale behind this step.

[GenoMax]

Sometimes if you are happy with the current state of the transcriptome (known expressed parts of the genome) then you could choose to do alignments of your data just to that part.

While that is not incorrect, you do run a small risk of having some reads mis-align (since an aligner does its best to align and the read may not have originally come from that region) by restricting to just "known" expressed parts of the genome. If splice sites are provided as well then the programs would not try to look for new ones. Both these modifications speed up the alignments to some extent.

[New2Bioinfo]

Both these modifications speed up the alignments to some extent.

That's okay. But while using the HISAT2 program, I am extracting the splice site and exon information from the .gtf file. And that information is given to the index builder (Hisat2-build). So, what I am getting is that this information during indexing is helping during alignment.

If I know the splice sites, the reads will not align to those parts where splice sites lie in the middle. Is this correct?
I still don't get how exon info is helping in alignment.

A little more detailed answer would be really really helpful.

Thank you.

[wdecoster]

The most intuitive explanation might be that those "known" exons and splice sites are used as a suggestion for the read mapping, making mapping much quicker since the aligner "knows" where to look. Reads that don't behave according to the "known" annotation will still get correctly aligned and new splice sites will be discovered.

You are just "telling" the aligner a priori where the splice junctions most likely are (but not restricting the mapping to those junctions/exons).

[New2Bioinfo]

Okay. That makes sense.

Thank you very much.

[biocomputer]

Does including exons and splice sites make the alignment more accurate, faster, or both?