It took us a while to figure out what was supposed to happen here, but the answer is that phenix.rna_validate is outdated and the numbers in MolProbity are correct. I'm fixing rna_validate now (and incorporating it into the new phenix.molprobity framework). -Nat On Tue, May 13, 2014 at 6:23 AM, Huw Jenkins wrote:

Hi,

I’ve recently been using phenix.erraser to help correct RNA geometry errors and finding it very useful. However, it appears that the geometry validation statistics reported by phenix.rna_validate with respect to bond angle outliers are very different from those reported by the MolProbity web server.

As an example - if I run phenix.rna_validate on 1u8d (as used in the documentation for phenix.erraser) the following outliers are reported:

RNA Validation ----------------------------------------------- Pucker Outliers: #residue:delta_angle:is_delta_outlier:epsilon_angle:is_epsilon_outler G A 15 :74.054:no:103.458:yes U A 48 :119.081:yes:283.965:no U A 63 :90.366:yes:74.220:yes

----------------------------------------------- Bond Length Outliers: #residue:atom_1:atom_2:num_sigmas

----------------------------------------------- Angle Outliers: #residue:atom_1:atom_2:atom_3:num_sigmas C A 39 : C5': C4': C3':-5.294 U A 63 : O3': C3': C2':-4.402

----------------------------------------------- Suite Outliers: #suiteID:suite:suiteness:triaged_angle G A 16 :!!:0.000:epsilon-1 U A 48 :!!:0.000:delta U A 49 :!!:0.000:delta-1 A A 64 :!!:0.000:epsilon-1

The MolProbity web server reports the same pucker and suite outliers but flags up many more angle violations:

A 39 C worst is C5'-C4'-C3': 10.588 σ A 63 U worst is O3'-C3'-C2': 8.803 σ A 54 C worst is C5'-C4'-C3': 7.883 σ A 49 U worst is C2'-C1'-N1: 5.416 σ A 62 G worst is C2'-C1'-N9: 4.946 σ A 15 G worst is O3'-C3'-C2': 4.825 σ A 22 U worst is C5'-C4'-C3': 4.785 σ A 64 A worst is C4'-C3'-O3': 4.648 σ A 65 A worst is C2'-C1'-N9: 4.582 σ A 44 A worst is C4'-C3'-C2': 4.395 σ

As the number of sigmas for the angle violations are different between phenix.rna_validate and MolProbity I guess they’re using different target distributions (not just reporting at a different sigma cut-off) but I’ve no idea which is more valid.

Hopefully there is a simple explanation!

Thanks,

Huw

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb