Genome vs transcriptome alignment

[rugilemat]

Hi,

I've been trying out m6anet on our RNA004 reads and I have a question about alignment. As a default, we tend to align our reads to the genome for isoform discovery. From what I noticed in the documentation, the recommendation seems to be aligning reads to the transcriptome rather than the genome.

I'm just wondering if aligning to the genome rather than the transcriptome makes a difference to the model's performance and would you recommend performing transcriptome alignments instead?

[yuukiiwa]

Hi @rugilemat,

We recommend to perform genome alignment and transcriptome alignment separately for isoform discovery and for m6A detection.

I suspect that if you input the genome-aligned eventalign.txt file to m6anet dataprep, it will not be able to output anything as the reference kmer will be the inverted model kmer.

Thanks!

Best wishes,
Yuk Kei

[rugilemat]

Thanks @yuukiiwa! I did get some output from genome-aligned bams, I'll see if I get to compare them with properly transcriptome-aligned files but I will stick to transcriptome-alignments for my analysis.

To see where the modifications occur in a particular transcript/gene would you then convert your m6a output to genomic coordinates as discussed in some other issues to be able to plot similar things to fig. 8 F in your preprint?

Also I'm just wondering if there's a particular number of iterations in –num_iterations you would suggest using with RNA004 model?

[yuukiiwa]

Hi @rugilemat,

Yes, I mapped the transcriptome coordinates to genome coordinates to plot Figure 8F.

I have been using --num_iterations 5, which works well with RNA004.

Thanks!

Best wishes,
Yuk Kei

[rugilemat]

Hi @yuukiiwa,
Thanks so much!

If I'm interested in read-level modifications from the documentation I understand that higher --num_iterations is recommended while --num_iterations 5 should be sufficient for site-level analysis?