Comparing RPKM across experiments

[Collidor]

Is it possible to compare RPKM values across studies? This question is something I have been thinking about during my own analyses. I tend to think that under the conditions of standard poly-A enrichment and approximately equal coverage during sequencing runs, then RPKM values for genes between separate studies can be compared. However, if study X used a transcript enrichment setup in which a select number of mRNAs from a subset of genes were being selected for sequencing, can the RPKM values obtained from this study be reliably compared to study Y in which poly-A enrichment was used?

[Collidor]

Part II for those who are feeling ambitious: If I were to selectively enrich for a set of transcripts (A,B,C), sequence these, and obtain high RPKM measurements and attempt to compare these to a poly-A enriched sample set in which I have a fair expression measurement of most of the transcriptome (although in this case transcripts A, B, and C were unaccounted for due to low abundance) is it possible to extrapolate the normal physiological mRNA abundance of these transcripts from the enriched data series?