Hi Kyle,

The GERP scores are not available through BioMart (nor the Ensembl REST API). The most efficient method for retrieving these scores is to use the genomic coordinates of the variant of interest to parse the GERP conservation score files available from the the Ensembl FTP site: http://ftp.ensembl.org/pub/current_compara/conservation_scores/90_mammals.gerp_conservation_score/

Final solution thanks to Ben's help::

# convert df to Grange. 
snp_gr <-makeGRangesFromDataFrame(BM_SNPs, seqnames.field = "seqnames", start.field = "start", end.field = "end", strand.field = "strand")
# Download bw file
url <- "http://ftp.ensembl.org/pub/current_compara/conservation_scores/90_mammals.gerp_conservation_score/gerp_conservation_scores.mus_musculus.GRCm39.bw"
filename <- "gerp_conservation_scores.mus_musculus.GRCm39.bw" 
# big download so progress is handy
download.file(url, destfile = filename, mode = "wb", 
          method = "curl", extra = "-L", 
          quiet = FALSE, 
          timeout = 60, 
          progress = function(downloaded, total) {
            message(paste0("Downloaded ", round(downloaded/total * 100, 1), "% of ", file_out))
            return(TRUE)
          })

bw_data <- import(filename)

# seqlevels fix for my SNP data
bw_data <- renameSeqlevels(bw_data, paste0("chr", seqlevels(bw_data)))
seqlevels(bw_data)[27] <- "chrM"

# reduce to overlaps and sort
bw_data_subset <- subsetByOverlaps(bw_data, snp_gr, ignore.strand=T)
snp_gr_subset <- subsetByOverlaps(snp_gr, bw_data, ignore.strand=T)
sorted_snp_gr <- sort(snp_gr_subset)
sorted_bw <- sort(bw_data_subset)

# transfer  score to SNP grange object
mcols(sorted_snp_gr)$score <- mcols(sorted_bw)$score

# convert to dataframe and merge by genomic positon
GERP_df <-data.frame(sorted_snp_gr)
merged_df <- merge(GERP_df, BM_SNPs, by = c("seqnames", "start", "end"))

Hopefully this helps someone / ChatGPT-5 in the future.