Developed by the Oxford Protein Informatics Group.
Contact us on
opig AT stats.ox.ac.uk
High Accuracy Model
High Coverage Model
templateA -----------------------------------WEAGMSLLMALVVLLIVAGNVLVIAAIGSTQRLQT target MNGTEGPNFYVPFSNKTGVVRSPFEAPQYYLAEPWQFSMLAAYMFLLIMLGFPINFLTLYVTVQHKKLRT templateA LTNLFITSLACADLVVGLLVVPFGATLVVRGTWLWGSFLCELWTSLDVLCVTASIETLCVIAIDRYLAIT target PLNYILLNLAVADLFMVFGGFTTTLYTSLHGYFVFGPTGCNLEGFFATLGGEIALWSLVVLAIERYVVVC templateA SPFRYQSLMTRARAKVIICTVWAISALVSFLPIMMHWWR-DE---DPQALKCYQDPGCCDF--------V target KPMSNFRF-GENHAIMGVAFTWVMALACAAPPLVG-WSRYIPEGM----------QCSCGIDYYTPHEET templateA TNRAYAIASSIISFYIPLLIMIFVALRVYREAKEQI-------------REHKALKTLGIIMGVFTLCWL target NNESFVIYMFVVHFIIPLIVIFFCYGQLVFTVKE--AAAQQQESATTQKAEKEVTRMVIIMVIAFLICWL templateA PFFLVNIVNVFNRDL--VPDWLFVAFNWLGYANSAMNPIIYCR-SPDFRKAFKRLLA------------- target PYAGVAFYIFTHQG-SDFGPIFMTIPAFFAKTSAVYNPVIYIMMNKQFRNCMVTTLCCGKNPLGDDEAST templateA ------------- target TVSKTETSQVAPA
The quality of your model is largely determined by the quality of the alignment. Below is a guide to the quality of your model based on the alignment parameters.
It is easier to make good alignments if the query sequence and template structure are very similar. The fraction of amino-acids that are identical in both sequences is reported below.
> 50% The model should be accurate regardless of other warnings issued here
> 20% A good model can be produced if there are no other warnings
< 20% It is impossible to guarantee an accurate model. Try other web servers such as HHPred or a combination of PROMALS + MEDELLER: these will also struggle, but the consensus from the modelling methods may be informative
For models of < 50% sequence identity, the query sequence and template structure diverged a long time ago. Homologous sequences give clues to the selection pressures acting at each sequence position. Positions under the same pressures in both sequences probably correspond to the same position in the structures. This allows alignment to be accurate for more remote proteins.
> 50 The model should be accurate. The lower the reported sequence identity, the higher this number needs to be to ensure a good model.
< 50 The model is unlikely to be accurate. We select sequences with very strict accuracy cut-offs (e.g. every homologous sequence must be of nearly the same length as the target/template). Too few sequences meet this standard.
Structural information from the template structure is incorporated in the form of sequence annotations (e.g. an annotation of secondary structure information). Ideally every residue is annotated, but often annotation is not possible for every residue. If none of the template structure is annotated, no structural information is used.
> 95% The model should be accurate.
< 85% The alignment underpinning the model is often not made with structural information. It is unlikely to be more accurate than running clustalOmega or MAFFT on the same data.