nearestTSS function in edgeR for differential methylation analysis

Question

[osa.zuta]

Hello,

I have hg19 Bismark mapped RRBS data that I would like to use edgeR on following the F1000 workflow (https://f1000research.com/articles/6-2055/v2 ). When I apply the function edgeR::nearestTSS to annotate my site-level data to compute the distance to the nearest gene's TSS, it appears to return hg38 based annotations. Looking at the nearestTSS code more closely, it seems to pull an object called CHRLOC from org.Hs.eg.db, but it isn't clear what reference genome (e.g. hg19 or hg38) that uses.

Main question: how does one interpret Distance output from nearestTSS if the reference genome is unspecified, and how does one specify it?

Thanks, zo

[osa.zuta]

Thank you for the quick answer. Any thoughts on how to best proceed with using nearestTSS on hg19 RRBS data?

Answer 1

[Gordon Smyth]

It is hg38. edgeR and the Bioconductor organism packages only support the latest official genome version for each species.

Later

Why not realign the sequence files to hg38? Why continue to use an old genome build? If you must use hg19 for some reason, then you could consider the TxDb.Hsapiens.UCSC.hg19.knownGene package. See the first version of the F1000Research article for how to use the TxDb package and Bioconductor genomic ranges functions instead of nearestTSS:

https://f1000research.com/articles/6-2055/v1

It is somewhat more complicated, which is why we switched to nearestTSS in version 2 of the article.