Gene-level rna2rna evaluation?

[rsuchecki]

When aligning RNA-Seq reads simulated from transcripts extracted from genome+annotation, many (unsurprisingly) align to an incorrect splice-form. Basic, rnf-style evaluation focuses on the exact position of the read. Should we, in addition to that, also consider capturing the numbers of reads aligning to the correct gene, as in some analyses this would suffice. You could also be more lenient about imprecise alignment within the correct transcript, but this would be relatively rare, so probably not worth pursuing.

[nathanhaigh]

This might be useful: http://dx.doi.org/10.1101/657874

[rsuchecki]

This might be useful: http://dx.doi.org/10.1101/657874

From the preprint:

[...] misquantifications, possibly arising from incorrect alignments, can impact DGE analysis, especially under conditions where the sequenced reads tend to diverge from the reference, such as in cancer and other disease conditions [...]
While we have focused on transcript-level analysis in the paper until now, here we look at differences in gene-level differential expression. This demonstrates that the quantification issues caused by lightweight-mapping or misalignment of reads can be of relevance even when one is performing gene-level analyses.