CASP-CAPRI collaborations


Recent progress in determining and predicting the structures of proteins and protein assemblies calls for closer integration of the different computational approaches for modeling assemblies and their building blocks. To enhance the integration between the different modeling techniques and between the distinct scientific communities that develop these techniques, CAPRI joint forces with CASP (Critical Assessment of Structure Prediction), whose traditional focus has been on the prediction of the 3D structure of individual proteins. Since 2014, CASP and CAPRI have conducted joint assembly prediction experiments during the regular CASP prediction season, organized every two years. For on-going or upcoming joint CASP-CAPRI assembly prediction challenges, see CAPRI Round announcements.

Please note that as in previous CASP-CAPRI collaborations, CASP assembly prediction targets where one of the components of the assembly needs to be modeled ab-initio (i.e. no templates are available in the PDB) will in general not be offered as target for the CAPRI prediction Round. Hence only a subset of the CASP assembly prediction targets will be designated for CAPRI.

Team registration

Teams can:

  • Participate in both CAPRI and CASP
  • Predictors have to register with both CASP ('season') and CAPRI ('round').
    When registering with CASP, participants should click the radio button "Do you want to receive an e-mail when a target is designated as CAPRI?" to receive automatic messages each time a new CAPRI target is designated. Participants should consult the CAPRI website (instead of CASP) for additional target information and for the timelines for submission of server models, docking models, and scoring submissions.
    When registering with CAPRI, participants have to provide their 12-digit CASP Team ID.
  • Participate only in CAPRI
  • Participants should register only with CAPRI, leaving the CASP ID field blank during registration. This option is offered for groups interested in the CAPRI scoring challenge only.
  • Participate only in CASP
  • Regular CASP participants should register only on the CASP website.

Only teams who provided their CASP Team ID during team registration on the CAPRI website are participating in both CAPRI and CASP. If no CASP team ID was given, the team’s submissions only go through CAPRI validator and will not be forwarded to CASP for evaluation.


As per usual CAPRI Rounds, participants are invited to submit 100 models in CAPRI format through the CAPRI gateway, with the first 10 models ranked in order of decreasing model quality. The CAPRI system will subject the submitted file through both CAPRI and CASP validation check. The CAPRI system then forwards the validated top 5 models to CASP for evaluation. Only models submitted through the CAPRI gateway will be considered for the scoring experiment.

Important notes

  • Model records have to be labelled correctly and organized in the correct order. Only the top 5 models will be converted to CASP format, and only if those models are labelled as MODEL 1 until MODEL 5. If Model 10 comes after MODEL 1 in the submission, only MODEL 1 is considered even if MODEL 2 comes after.
  • Participants can add a PARENT record to specify the structure template(s) used to generate each model. This record is inserted in the line immediately following a MODEL record. If no PARENT record is specified, the CAPRI system automatically assigns "PARENT N/A".
       Indicates that a prediction is not directly based on any known
       structure. Note that this is the only indication in the file that the
       prediction is ab initio, so is a critical piece of information.
     PARENT 1abc_A
       Indicates that a single PDB entry 1abc, chain A 
       was used as a modeling template.
       All template-based predictions should be submitted with this form 
       of the PARENT record. Note that, in order to be accepted, the code 
       must correspond to a current PDB entry.
     PARENT 1cdc 2def_g [3hij_k ...]
       Indicates that the model is based on more than one structural template. 
       Up to five PDB chains may be listed here with additional detailed information 
       included in the METHOD records. 

Common errors during CASP format validation

This section aims to list common errors we have seen for CASP validation based on previous CASP-CAPRI collaborations.

  • # ERROR! Sequence of chain ? in model doesn't match any sequence of subunits.
    • Usually caused by non-canonical residues. Suggestion:
      Change HSD/HSE/HIE/HID/HIP to HIS
      Change CYX/CYM to CYS
      Change GLH to GLU, HYP to PRO, MSE to MET, ASH to ASP
    • Your residue sequence numbering does not match the template given (column 23-26 of ATOM records).
  • # ERROR! Unacceptable index of the prediction MODEL
  • MODEL record doesn't specify model number correctly. E.g MODEL 001 instead of 1.
  • # ERROR! Check residue: I ? chain ? (! atom CD INCONSISTENT with residue ILE !)
  • Please change atom CD to atom CD1 for ILE before resubmitting.
  • # ERROR! Check residue: ? ? chain ? (! There is no CA atom or its occupancy equals 0.0 !)
  • Please make sure CA atom is present for each residue before resubmitting.

Past CASP-CAPRI collaboration (published reports)

  • CASP13-CAPRI Assembly prediction; CAPRI Round 46
    Blind prediction of homo- and hetero-protein complexes: The CASP13-CAPRI experiment. Lensink MF et al. Proteins (2019) 87(12):1200-1221; PMID:31612567
  • CASP12-CAPRI assembly prediction; CAPRI Round 37
    The challenge of modeling protein assemblies: the CASP12-CAPRI experiment. Lensink MF et al. Proteins (2018) Mar;86 Suppl 1:257-273; PMID:29127686
  • CASP11-CAPRI assembly prediction; CAPRI Round 30
    Prediction of homoprotein and heteroprotein complexes by protein docking and template-based modeling: A CASP-CAPRI experiment. Lensink MF. et. al. Proteins. 2016 Sep;84 Suppl 1:323-48; PMID:27122118