[BioC] multi-mapped reads: Cufflinks + baySeq? edgeR?

[Hilary [guest]]

Hi,
I am trying to analyze RNA-Seq data for (gene-level) differential expression between treatments, in a design incorporating multiple factors (effects of species * treatment & interaction, 4 replicates for each combination). I have reads that map to multiple locations (Single-End data) and while I'd first used Bowtie2/Tophat >> htseq (discarding multi-mapping reads = multihits in htseq), and then used the GLM and baySeq approaches, it was suggested I go back and include multi-mapping hits. 

I know Cufflinks allows incorporation of the multi-mapping reads (Mortazavi method I think), and I know it is not compatible with the GLM methods of edgeR/DEseq due to use of FPKM but does that incompatibility apply to baySeq as well?

Using CuffDiff seems problematic as it only does pairwise tests -- and while I can do that, I think a full model testing for individual effects and their interaction (esp. as our real interest is the species*treatment interaction) is probably more statistically accurate. Thus I'm not sure how to proceed; any suggestions would be greatly appreciated if someone has time! 
Thank you,
Hilary

 -- output of sessionInfo(): 

()

--
Sent via the guest posting facility at bioconductor.org.