Model Evaluation Help Pages







Job Name

The Job Name is used to identify the status in the queue and to display the results.

Modeller Key

To use the evaluation server, input of the MODELLER license key is required. For further information, please consult

Email Address

An email will be sent to the input email address, once the calculations are finished.

Model File

Input model file in PDB format (see

Alignment File

Optional input alignment file in PIR format (see MODELLER manual). Please note that the second field in line two of the structure entry should contain a pdb code (without chain) of a template file currently available in the PDB database.

The TSVMod score is significantly more reliable if:

  • TSVMod receives a model and an alignment file
  • The template PDB code or a closely related template can be found in the TSVMod training set

Sequence Identity

The calculation of the GA341 score requires the target-template sequence identity. For model files produced by MODELLER, this information will be taken from the PDB header, so this field can be left blank.



TSVMod predicts the RMSD and the Native Overlap (at 3.5 Å cutoff) of a model with its native structure, using a support vector machine (SVM).

Predicted RMSD


The predicted root-mean-squared deviation (RMSD) between the coordinates of the Cα atoms in the model and in the native structure.

Predicted Native Overlap

Fraction of model Cα atoms that are predicted to be within 3.5 Å of their positions in the native structure.

Match Type

  • MatchByTemplate: Training set is based on the template structure, or a closely related template structure (95% sequence identity). This match type returns the most reliable results. It can only be used with an input alignment file in MODELLER-style PIR format, and if the template has been used in the training set.
  • MatchBySS: Training set is based on similar secondary structure features. This match type is used in two ways: with all features, and with a reduced feature set. If an alignment file is given, this match type is used if the template PDB file can not be found in the training set, and the information of templates with similar features is used instead. Without an alignment file, only a reduced set of features is used in the SVM, resulting in the least reliable prediction.


  • All: all available z-scores are used in the SVM
  • Reduced: for lack of sufficiently similar structures in the training set, a reduced set of z-scores is used to compute the results.

Relax Count

The standard relax count is 1. If not enough similar structures are found in the training set, the default boundaries are relaxed incrementally.

Set Size

Number of similar structures used to compute the results




Using probability theory, we derive an atomic distance- dependent statistical potential from a sample of native structures that does not depend on any adjustable parameters (Discrete Optimized Protein Energy, or DOPE). DOPE is based on an improved reference state that corresponds to noninteracting atoms in a homogeneous sphere with the radius dependent on a sample native structure; it thus accounts for the finite and spherical shape of the native structures. The evaluation server reports the normalized z-DOPE score.

DOPE Profile

DOPE energy profile, smoothed over a 15 residue window, normalized over the number of DOPE restraints acting on each residue.




Score for the reliability of a model, derived from statistical potentials (F. Melo, R. Sanchez, A. Sali, 2002). A model is predicted to be reliable when the model score is higher than a pre-specified cutoff (0.7). A reliable model has a probability of the correct fold that is larger than 95%. A fold is correct when at least 30% of its Cα atoms superpose within 3.5Å of their correct positions.


A surface statistical potential that contributes to GA341.


A pairwise statistical potential that contributes to GA341.


A combined statistical potential that contributes to GA341.