Problem in nucleotide-protein interface scoring

[lafita]

Some nucleotide-protein interfaces have a very low score, like interfaces 6 and 7 of 5B1L.

I suggest to do a more thorough analysis to see if this is a common case, because here the DNA is rolled around the protein complex, making pairwise interfaces weaker compared to the whole complex interaction. If this is common (to obtain almost zero scores for protein-nucleotide interfaces), we could treat them like nucleotide only interfaces with a NOPRED and 0.5 probability.

This is also a special case because it is a big heteromer and the reported assemblies are constructed using heuristics, so the assembly which would have the highest probability (the co-crystal of the DNA and the protein complex) is not reported. Therefore all assemblies have a very small probability.

[josemduarte]

#161 is related to this