Gene map ensembl ids are transcripts probeToMeanPromoterMethylation

[FischerJoBio]

The returned Ensembl ids of the probe map of probeToMeanPromoterMethylation are transcripts (ENST*) but should be genes (ENSG*), as probes are mapped to the genome and not transcript-level information

[katehoffshutta]

This behavior actually comes from the generation of the probe map in mapProbesToGenes. The rationale for doing the mapping at the transcript level is to identify probes that are within a certain range of the TSS for any of the various transcripts of a gene that are recorded in the transcriptIDs column of the Illumina manifest. We actually have to map to the transcript level for this reason, I think.

[FischerJoBio]

I see your point, but this is problematic for most downstream applications (including everything we do), as we work on a gene-level (i.e., reference genome annotation) and not transcript-level. Then the question becomes, how do we map from the transcripts to genome-level? Also, these methylation counts are a genomic feature and not mRNA feature. There is reasoning for both ways, but we should decide and keep it consistent for practicality -- and if we stick to transcript-level promoter definitions we should warn users and provide a function to map to the genome.

[katehoffshutta]

Based on the manifest use of transcripts vs. genes, I propose this resolution @FischerJoBio:

• Add a note that promoter definitions are at the transcript level
• Advise users that they can write their own functions to pre-filter the input probe_gene_map if they are looking to calculate promoter methylation for a single transcript per gene

I don't think we need to provide a function to map to the genome at this point, but we could open it as a separate issue for enhancement if desired.

[katehoffshutta]

Note that when we do the mapping in probeToMeanPromoterMethylation, the current behavior is to include a probe multiple times if it maps to multiple isoforms. This also needs to be included in the documentation. MWE:

library(NetSciDataCompanion)
genesInt <- c("NLGN4X") #,"MAP7D2","SH3KBP1"  ,"MAP3K15","ARSD" , "ARSF" , "MXRA5"  ,"NLGN4X"  ,"PUDP")
objNet2 <- CreateNetSciDataCompanionObject()
probeList = c("cg17811446","cg20303283","cg21875437",
              "cg21142738","cg06811375","cg08881159",
              "cg15790913","cg19743317")
probemap <- objNet2$mapProbesToGenes(probeList,rangeUp=500, rangeDown=0)

set.seed(1989)
betas = rbeta(n=length(probeList),shape1=0.5,shape2=0.5)

gene_prom_meth <- objNet2$probeToMeanPromoterMethylation(methylation_betas = data.frame("probeID"=probeList,"person1"=betas),
                                                         genesOfInterest = genesInt,
                                                         probe_gene_map = probemap)
gene_prom_meth
# 0.45119
mean(betas)
# 0.47121
1/9*(sum(betas) + betas[which(probemap$geneNames == "NLGN4X;NLGN4X")])
# 0.45119