I am trying to put together a dataset of hg38 human genome SNP data (CDS only) that contains information on:

• SNP chromosomal location
• Allele frequencies
• Allele change
• Codon change
• Peptide change
• Variant consequence
• Gene within which SNP is located

Initially I tried to use BiomaRt's getBM to get variation data, which included everything I wanted except allele frequencies (you can get minor allele frequencies but only very few of the entries actually have any data for this).

I eventually gave up on BiomaRt for variation data as it keeps crashing R (I know my code is correct because it works on the smaller chromosomes - 22 for instance - but fails on the larger ones). From reading other threads I can see that the variation data on BiomaRt can be fiddly to acquire due to the size of the datasets.

So my new approach is to acquire data via the UCSC table browser which gives me all the information above except the gene within which the SNP is located. I have managed to get gene information from BiomaRt using:

ensembl = useMart("ensembl",dataset="hsapiens_gene_ensembl")
attributes = listAttributes(ensembl)
#Get all chromosome names
chroms <- getBM(attributes = 'chromosome_name',
                mart = ensembl)

for(i in 1:length(chroms[,])){
  gene.table <- getBM(attributes = c('ensembl_gene_id', 'external_gene_name', 'chromosome_name', 
                                     'ensembl_transcript_id', 'transcript_start', 'transcript_end', 
                                     'exon_chrom_start', 'exon_chrom_end', 'phase', 'end_phase', 
                                     'genomic_coding_start', 'genomic_coding_end', 'cds_start', 'cds_end'), 
                      filters = c('chromosome_name'),
                      values = chroms[i,1], 
                      mart = ensembl)

  gene.table <- data.frame(t(sapply(gene.table,c)))
  gene.table <- t(gene.table)
  gene.table <- gene.table[complete.cases(gene.table),]

  write.table(gene.table, file=paste("chrom.",chroms[i,1],".gene.txt", sep=''), sep='\t', quote=FALSE)
}

From here I can use the SNP chromosomal location and the gene_coding_start and gene_coding_end coordinates to identify to which gene each SNP belongs.

The reason for my post is that this seems like a rather unwieldy approach to acquiring what I assume is a commonly used dataset. What I would like to know is whether there are more efficient ways of getting what I need, and if not is my own methodology sound.

As an aside, I did intend to use VEP but have had real issues attempting to install it or use it via the virtual machine.