What is the difference between the I-TASSER, phyre2, and SWISS-model in the 3D tertiary structure?

How do they get the results? When I did a prediction, I got a similar result for the highest template in I-TASSER and Swiss, while it was different with Phyre2. What does that mean?


1 Answer 1


This is a common Q.

Briefly, you have some techniques here that get used separately or combined:

  • Ab initio: assemble the protein from scratch, often via a library fragmented peptides (3-mer, 6-mers, 9-mers) based on real structures.
  • Template threading: given a protein structure alter it to make it like a give sequence. SWISS-MODEL, Rosetta's SimpleThreadingMover, ModBase, Phyre template threading etc. This is why some of these keep ligands and deal well with multimers. For missing loops in the template these are filled by ab initio methods.
  • Hybridisation: give a bunch of models mix and match them for the best energy. I-Tasser, RosettaCM (which uses SimpleThreadingMover), Phyre (which uses Phyre template threading)
  • Ab initio with contraints, such as evolutionary covariance. EVFold, AlphaFold2, trRosetta. Often mixed with template based approaches.

Wikipedia has more, but papers are very helpful despite the car-salesmanship.

One thing to note is that protein can adopt different conformations. The AlphaFold2/ColabFold notebooks give multiple models: these may be legit alt conformations, e.g. inactive vs active etc etc

Here is a SwissModel vs AF2: Different structures of same protein from two softwares

Other helpful As:


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.