I have not seen examples of restriction recognition sequences stored as GFF3. However, I'll take an educated guess!

First, I think that it makes sense for start/end to refer to the palindromic sequence. In other words as we discussed above for EcoRI, start = 1, end = 6. The cleavage site could be stored in the last column (as a free text attribute or note).

Second, in terms of storing the fragments that result from digestion - I would not. But if I did, I would think of them in terms of only one strand. So for example, again using EcoRI:

G^AATTC

The fragments would be represented by 2 features: one of length 1 (G), the other of length 5 (AATTC). I guess they would best be described as children of the parent feature (the uncleaved site) ? Would be interested in other opinions on that.

It's probably worth searching the SO using the term "restriction" to get a good sense of how sites, fragments etc. are defined.

I'm not aware of a standard for this. It's a bit complex given that restriction enzymes can have multiple cleavage sites, some of which may be outside of the recognition site. e.g.

HaeIV       (7/13)GAYNNNNNRTC(14/9)

in REBASE notation.

I would use multiple, linked GFF features; one for the recognition site, others for the cleavage site(s). Putting the REBASE notation into the feature tags wouldn't hurt.

Most annotation systems, including GFF3, generally ignore the issue of DNA strandedness for the main sequence. That said, there is the SO:0000984 (single strand) sequence attribute that can be applied to located sequence features. There isn't a specific SO term for sticky ends. For your digested fragments, I would use such features to indicate the sticky ends.

If you have a lot of annotation that will become public, I would contact the SO curators to discuss some specific new terms because there don't seem to be any in SO at the moment e.g. RE recogition site, sticky end etc.

If you put the nomenclature of Rebase in your file to define the restriction site (for example EcoR1 is defined as:

G^AATTC

) you should be able to get the position of the sticky ends one your sequence on both strands.

For example:

cut=    {
    enzyme:"EcorRI",
    site: "G^AATTC", ## cut position are 1 and length()-1=5
    position: 1
    }

then on the left side of the cutting site, the sequence ends at 1+1 on the "+" strand and ends at 1+5 on the "-" strand.

for the second part of your question, you should only consider the "+" strand . The sticky site won't be extended unless you use a DNA polymerase. ( e.g. T7 DNA polymerase has a 3' -> 5' exonuclease activity. )

Here's what I do:

16      xxx     restriction_enzyme_binding_site 69742532        69742536        .       .       .       enzyme=FatI;ID=4ca20b5cf9d96d3ddb000000;site=CATG
16      xxx     restriction_enzyme_binding_site 69743737        69743741        .       .       .       enzyme=FatI;ID=4ca20b5cf9d96d3ddb000001;site=CATG
16      xxx     restriction_fragment    69742532        69743741        .       .       .       enzymes=FatI;left_site=4ca20b5cf9d96d3ddb000000;ID=4ca20b5cf9d96d3ddb000059;right_site=4ca20b5cf9d96d3ddb000001

Site positions match the entire palindromic sequence, fragments extend over the single stranded DNA.

But then FatI is not that complicated.

I believe in GFF3 you can use a hierarchical approach. Therefore, list the target site in one line.

Here's my "pseudogff" which lists the restriction site. Then, the start/stop as position 2 for the cut site.

genomeId source ..... name=G^AATTC;seqid=cutSite1
cutSite1 source .. 2  2 ......