clarify bed versus bedgraph format

[cokelaer]

Once done, we can fix other issues such as bam2bed, rename bedgraph into bed or vice versa etc.

As also pointed in this issue from sequana_coverage, sequana/sequana#555
the BED format is made of 4 columns: chrom name, start pos end pos, value.

The difference between the BED and BEDGRAPH is that BEDGRAPH is 4 columns whereas BED can be up to 12 columns.

In bam2bedgraph, bedtools genomecov -bg -ibam creates a bedgraph*:

chr1	75	176	1
chr1	447	547	1
chr1	547	548	2
chr1	548	648	1
chr1	661	690	1

In bam2bed each position (even those with no coverage) are reported with bedtools genomecov -d
and therefore the end position is not reported. we get for instance::

chr1	1	0
chr1	2	0
chr1	3	0
...
chr1	75	1
chr1	75	1

but this is not a BED file. Actually, this rather a COV (for coverage) format, but there is no such standard. This is also the same output given by samtools depth .

[cokelaer]

We introduce a new format to store the coverage. The class bam2bed is now called bam2cov to avoid ambiguity. The bam2bedgraph is still available.

[blaiseli]

the BED format is made of 4 columns: chrom name, start pos end pos, value

Sorry for this late comment: Note that in the BED format, 4-th column, if present, should be a name, not a value, which makes it different from BEDGRAPH

According to https://genome.ucsc.edu/FAQ/FAQformat.html#format1:

The order of the optional fields is binding: lower-numbered fields must always be populated if higher-numbered fields are used.

And the order is chrom, chromStart, chromEnd, [name, score, strand, ...]

When I need something BED-compliant, with strand information, I sometimes have to introduce dummy values for name and score. If I want something with a score, I sometimes need to introduce dummy values for name.

And, if we really strictly want to have a BED-compliant format, the score, if present, should be between 0 and 1000 (same reference as above).