methylartist: ValueError: Could not interpret value `coord` for parameter `x`

Ignore the above comment, I’ve modified the arguments to composite in commit 13522b7 so that this makes some sense now I think. There’s now a --minelts and --maxelts which default to 1 and 200, respectively. To plot everything you can just give --maxelts a very large value e.g. 5000 e.g.:

methylartist composite -d 054.data.txt -s te/SVA.1kbp.bed -r Homo_sapiens_assembly38.fasta -t SVA_F.fa -p 64 --output_table --lenfrac 0.5 --maxelts 5000 -a 0.1

Also, this now shows a message e.g. 2022-09-29 23:36:24,950 sample 054 has 1563 useable elements, will sample 1563 which can be compared against the number of unique elements in the output table for a sanity check e.g.:

$ cut -f1,2,3 054.data.SVA.1kbp.composite.table.tsv | grep -v ^chrom | sort | uniq | wc -l
1563

I’ve added a --output_table option to composite in commit 737a5c0. The output currently looks something like:

chrom   start   end     strand  sample  coord   meth
chr16   24108350        24110141        -       054     33      0.6178728620349763
chr2    202534282       202536111       -       054     33      0.90873449102059
chr20   32723118        32724962        +       054     33      0.9506836249230268
... etc ...

where coord refers to a position in the reference element (.fasta passed to -t option). The positions included in the table are determined the same way the positions included in the mean methylation plot (default to median coverage or --meanplot_cutoff. Let me know if this works for you.

Good to know it works for you. It should be straightforward to dump the elements x sites data into a table, I’ll look at it. It will be a rather sparse table if all CpGs for all elements are included but that’s probably OK. Alternately could use the same --meanplot_cutoff as above.

Also, it’s probably worth noting that composite samples from the list of input sites up to --maxsegs elements, so if you want to see everything plotted this must be set to a number at least as big as your input.

I see similar results for SVA, it was due to their heterogeneity and an automatic cutoff that determines which sites are included in the mean methylation plot. Because each individual element (in this case SVA) might include different sites (in this case, CG dinucleotides), each position in the reference SVA aligned to the individual SVA in is assessed to see how many SVAs have a CG at that position, call this the “coverage” at that CG. The median coverage is taken across all CGs and used as a cutoff for determining which to include in the plot. Since SVAs are very heterogeneous this doesn’t work well so I’ve added a parameter --meanplot_cutoff in f6f7fee, let me know if it works for you, looks like this for me (default vs --meanplot_cutoff 5):

default:

--meanplot_cutoff 5: