There seems to be a need for more complex ways to detect/filter
    overlaps of RNA read alignments with transcript annotations. These
    could be implemented as new overlap types and/or as ways of
    filtering RangesMatching objects.

    Different types of filters:
    * Compatible splicing, i.e., equal gaps
    * Compatible on both ends, for paired end data
    * Incompatible (inverse of above)
    * Retained introns (read covers an intron, without a splice (gap) in the
      same intron, i.e., read completely covers intron or a read end
      starts/ends in the intron)

    It would seem the above would benefit somehow from a way of
    representing paired end, spliced reads. GRangesList is close, but
    there is no way to distinguish the insert gaps from the
    others. Maybe the gap could be stored in an IRanges column in the
    element metadata?

    While things like countOverlaps and summarizeGenomicOverlaps
    support a number of useful counting methods, it would be useful to
    have some of those on RangesMatching, for the sake of efficiency.
    Possible counting methods on RangesMatching:
    * Counting number of hits for each query/subject (query count is
      already available via as.table)
    * Counting number of hits for each uniquely mapping query

    At a slightly higher level, aggregating these (per transcript)
    counts to the gene level (this is not a simple table(), because a
    read should only be counted once for each gene, even if it hits
    multiple transcripts).

    Also, it would be useful to have set-style (union, intersect,
    setdiff) operations on RangesMatching.

Just putting this out here for discussion and maybe we (Herve?) can start
working on at least the first suggestion, checking for splicing
compatibility.

Thanks,
Michael

	[[alternative HTML version deleted]]