On Fri, Oct 12, 2012 at 6:54 AM, Edward A. Berry
From a recent PDB validation: ==> The following table contains a list of torsion angles outside the expected Ramachandran regions [GJ. Kleywegt and TA. Jones, PHI/PSI-chology: Ramachandran Revisited, Structure 1996, 4, 1395 - 1400].
Residue Chain Sequence PSI PHI ASN A 35 -101.76 -94.29 LYS A 45 -111.15 60.76 LEU B 12 84.56 -68.33 ASN B 35 -102.05 -94.34 LYS B 45 -127.20 70.13
Looking at Kleywegt's paper, their allowed region contains 98% of residues in hi-res structures. So if your protein has, say 1000 residues, you would expect about 20 true "outliers".
This may be appropriate for the PDB validation - for Phenix/Molprobity, where the meaning of "outlier" is probably different, the guideline we use is a maximum of 0.2% outliers, with 98% in "favored" regions. -Nat