Martin Hallberg wrote:
I recently received a referee report that stated "The authors should
describe the number of residues in the fully allowed, favored. The
present statistic is misleading. Molprobity is far too loose (it even
gives good values for structures that are wrong!) Procheck is more
realistic."
That is basically the inverse of my own experience and there are some
posts on the CCP4BB supporting this view (see for example http://
www.dl.ac.uk/list-archive-public/ccp4bb/2005-07/msg00227.html ). My
first reaction was that the reviewer is simply not up to date. After
checking a good number of recent papers, I find that almost all papers
(of which some are by people active on this board) refer to the
Procheck definition. Is this because Procheck is still distributed and
supported by the CCP4?
People know and are familiar with MolProbity.
I suggest you send the quote from the Molprobity manual back to the
referees on this one. I would also cite its use by the PDB. You may also
quote me if you like:
"The ProCheck Ramachandran plot data is based on old data, and it is my
understanding that the software is no longer updated by the author.
MolProbity by contrast includes new data from a large number of high
resolution structures solved since the introduction of ProCheck and
refined using more modern algorithms. In any disagreement between
MolProbity and ProCheck, I would consider the MolProbity output to be
authoritative. Kevin Cowtan, CCP4 bulletin board, February 2007"
