search

rcsb / symmetry-benchmark

Benchmarks for testing the symmetry package
3 stars 8 forks source link

d1em8c_ #25

Open lafita opened 9 years ago

lafita commented 9 years ago

Annotated as C1, but contains a clear C2 symmetry with two helices (one large and one small) and two strands (one of them is swapped into the other subunit) per subunit. SCOP does not say anything about duplication or symmetry, but the domain is annotated as alpha-beta-alpha. It is an interesting case that should be considered to tag as C2.

img

d1em8c_bmerror

http://scop.berkeley.edu/astral/jmolview/?ver=2.05&id=d1em8c_&context=0

lafita commented 9 years ago

Comments on the original benchmark manual validation: it has babab motif, but protodomains aren't symmetric. Asymmetric insertion sites on cesymm alignment and possible strand rearrangement. The protodomains are symmetric even taking into account the insertions and strand rearrangements.

sbliven commented 9 years ago

Interesting. Definitely a borderline case. The order of b-strands in the middle is odd–7 1 6 5 2 4 3. It's true that omitting strand 5 from the alignment makes a decent alignment without resorting to strand swapping, but it's weird to treat such a fundamental part of the structure as an insertion. From an evolutionary perspective, it's clear that this couldn't arise from a simple duplication, but would have had to undergo some significant structural rearrangements in the core. I think that's why we settled on C1, although from a structural point of view the C2 seems decent enough.

lafita commented 9 years ago

The point is if we are also considering evolutionary criteria for the symmetry classification or we stick only to the structural similarity between the parts.

To me the evolutionary reasoning seems a task that user has to do when the structural results are given, because algorithmically it is very difficult to decide if the parts not included in the symmetry are relevant for an evolutionary event or not (the only variable that would change is the coverage).

This example seems more like a convergent evolution of the two parts into a symmetric fold, rather than a duplication event, but the overall domain is symmetric in any of the two cases, so we should tag it as C2 to be consistent.

andreasprlic commented 9 years ago

I am slightly concerned about using CE-symm to re-evaluate the benchmark. The benchmark should really be independent of any algorithm. As such I'd vote for keeping it C1, even if I see that this is a borderline case.