Method:
The keyword "transferase" was first used to search PDB for transferase proteins with known structure. One protein was picked out from the hits to
run blasp service against sequence databases. Then Xenopus laevis Glutathione S-transferase was chosen from the hits which satisfying the requirement.
(length=212aa,identy between 50% and 60% to known structures) The structure of Xenopus laevis Glutathione S-transferase was predicted by two different
methods.One method was homology modelling by tool SWISS-MODEL.I used "First Approach Mode" in the Swiss-Model server. Structure templates were selected by homology from ExNRL-3Ddatabase to do the modelling.The other method was fold recognition by 3D-PSSM. The template having greatest confidence, with PDBid 6gsv was used to model the structure.The templates used in both methods were totally different.

---

Result:
The quality of modelling was evaluated by tool WhatIf Check for both structures. In all, 6 errors and 15 warning were found for the structure predicted by
SWISS-MODEL.
Validation for structure by SWISS-MODEL
4 erros and 18 warnings were found for the structure modelled by 3D-PSSM.
Validation for structure by 3D-PSSM
To see the difference of these two modellings,CE structural alignment was performed for these two structures.
From the result, it can be concluded that the overall structure prediction by both methods agree well, however small discrapancy lies in both backbone and sidechain configurations.