The Model Overview section of the NovaFold Report allows you compare the structural similarity of a NovaFold prediction to experimental structures in the RCSB Protein Data Bank (PDB).

  • Model – The model ID number.
  • TM error – The error (± value) for the TM-score calculation.
  • RMSD error – The error (± value) for the RMSD calculation.
  • C-score - Confidence score [-5,2], with higher numbers denoting increased confidence.
  • Cluster size – During the “ab initio” simulations performed by NovaFold, thousands of plausible structure conformations are recorded. These conformations are clustered and one representative from each cluster is selected as the scaffold that is used to create the final prediction(s). Cluster size is the number of conformations in the cluster. A larger Cluster size is indicative of a lower-energy conformation. The lowest-energy conformations are associated with biologically-relevant structures. Therefore, with regard to determining ranking, a higher Cluster size is better.
  • Density score - This score is related to the tightness of the cluster and similarity of the cluster to the threading templates. It is calculated as:

Density score = Ncluster / (Ntotal x RMSD)


    • Ncluster is the number of recorded conformations assigned to the cluster that the final model represents.

    • Ntotal is the total number of recorded conformations from the simulation process.

    • RMSD is the average root-mean-squared deviation of the cluster’s conformations to the “centroid” cluster representative. The minimum value is 0.8 to prevent a division-by-zero singularity.

Density scores range from 0 to 1.25. A higher score is better with regard to determining ranking.


TM-score, TM error, RMSD and RMSD error are color-coded to indicate the predicted quality of the resulting models.

  • GreenTM-score ≥ 0.75. Models are appropriate for drug design, drug screening, ligand docking, and molecular replacement.
  • Orange - 0.50 ≤ TM-score < 0.75. Models are appropriate for mutagenesis design and predicting active sites, disease substitutions, alternative splicing patterns, and domain boundaries.
  • Red - 0.17 ≤ TM-score < 0.50. Models might be appropriate for predicting domain boundaries.
  • BlackTM-score < 0.17. Similarity between models can occur at random.

Below this section of the NovaFold Report are sections for each individual model.

Need more help with this?

Thanks for your feedback.