[feature request] dumping color file into parsable format

[damiankao]

It would be great if we can dump the color file into a text format or something easier to parse so we can analyze the graph.

[rhysnewell]

I would also appreciate this, alongside the index file if possible. Or perhaps some instruction on how to interpret the byte strings in the colours file?

[cgroza]

This would be useful for creating phylogenies based on k-mer sharing between colours using metrics like Jaccard distance.

[cgroza]

The right solution to this is to use the Bifrost API. I have made an attempt here, maybe it is useful to others: https://github.com/cgroza/bifrost_jaccard

[GuillaumeHolley]

Hi everyone,

@cgroza Thank you for your implementation!

There is now another fairly simple solution to do this:

1. Make a FASTA file of each k-mer in the segments of the GFA file. Assuming k=31:

   zcat mygraph.gfa.gz | awk 'BEGIN {K=31} {if ($1=="S"){LEN_KM_UNITIG=length($3)-K+1; for (i=1; i<=LEN_KM_UNITIG; i+=1){print ">" $2 "_" i "\n" substr($3,i,K)}}}' > mygraph.kmers.fasta

   Every record in the generated FASTA file has a name with the form >x_y where x is the unitig ID (in the GFA) the k-mer is from and y is the position (1-based) of the k-mer within that unitig.

2. Query the colored graph using the previously generated GFA

   Bifrost query -v -t 16 -e 1.0 -g mygraph.gfa.gz -C mygraph.color.bfg -q mygraph.kmers.fasta -o mygraph.colors

   The output file will be mygraph.colors.tsv which is a matrix (k-mers x colors). The intersection of a row (k-mer) and column (color) contains a binary value indicating whether the corresponding k-mer is present (1) or not (0) in the sample matching the corresponding color.

Given that Bifrost graphs are now fully indexed (.bfi file output alongside the .gfa) and are very fast to load in memory, this solution should take no time to run.