mflamand
commented
2 years ago Hi, The short answer is that it does not. I have kept the parsing of bam files to be done in a similar way that was done in the hyperTRIBE pipeline, which was the inspiration for Bullseye. I have thought of ways to this and may implement this in the future. We simply never had any issues with the way its working now. Let me explain in more detail.
We prepare our libraries using a stranded library prep method, which should result in a good ratio of reads on the correct strand. Bullseye uses samtools view to parse all reads mapped in the genome and construct a matrix of coverage for each position with the number of each nucleotide. Samtools outputs all reads in the same direction (sense), thus this parsing is not stranded. It should be straightforward to split reads based on orientation and add information on strand for each position. This would increase the matrix file size by less than twice. The "strand" decision is taken when the refFlat file is provided. For each transcript in this file, it builds a list of genomic interval for features: CDS, 3UTR, 5UTR, ncRNA, introns, etc depending on option used. and does this for both positive and negative strand. It than looks within these defined intervals how many of each nucleotide are mapped. On the negative strand it reverse complements the coverage. So we are still looking at C-to-U (C-to-T) transitions on either strand, but only when a transcript is annotated at a given position. This can be problematic if 2 RNAs on opposite strands are overlapping. In my experience I did not find instances of sites being found on 2 overlapping highly expressed RNAs. In any case the C-to-T transition can only be in one direction. For example if there is a mutation C-to-T transition on the reverse strand, on an overlapping RNA on the positive strand, we would see an a G-to-A and it would not be called. Of course in this case, the actual frequency of edits would be skewed because we did not keep track of the original strandedness of the read. But as I said these are rare cases. I nonetheless think that keeping track of strand would be an interesting feature to implement.
Does that make sense? Let me know if you have further questions.
ckuenne
hi, this might be a bit of a stupid question: but how does bullseye handle strandedness of the library (first stranded / second strand)? see here: https://chipster.csc.fi/manual/library-type-summary.html