13 months ago by
I'm not sure if this exactly answers your question, but:
1) Some conversions between expression measurements distributing multi-mapped reads should be possible, but I would have to double-check exactly what was provided to see if I could answer your question about the TPM conversion.
Also, in Xena, I believe you usually search by gene symbol (not transcript ID). So, I'm guessing transcript counts have already been summed by gene symbol (or multi-mapped reads were estimated between genes, rather than among transcripts for the same gene?).
2) The UCSC Xena browser actually provides two different expression measurements, but it is not possible to convert between them (one uses RSEM with multi-mapped read distribution, the other uses htseq-count for only unique reads). However, if you want to get some sense of robustness in your results, you can test both sets of results (with and without "GDC" in the name).