Developing an expert system to improve the prediction of the tridimensional protein structure

Project: PHD

Project Details


The aim of our work is to propose a reliable methodology for homology
modeling, especially when the protein of interest shares a low
percentage of identities (20-30%) with the template.

Similar sequences are fetched (PSI-BLAST[1]) in a non-redundant sequence
databank. Then, as far as possible, two sets of sequences are built.
This method aims at creating different conditions to run multiple
alignment programs and extracting different consensus and in order to
raise the confidence of the sequence-structure alignment.

The two sets are then submitted to alignment programs, ClustalW[7],
Dialign2[5], Match-Box[3], Multalin[2] and PRRP [4]. A pairwise
alignment between the target and template sequences is extracted from
each multiple alignment and the final sequence-structure alignment is
obtained from the consensus between all the pairwise alignments. A
tri-dimensional model is built using MODELLER[6] on this final
alignment. As a control, another model is built from the rough
sequence-structure alignment provided by PSI-BLAST[1], and compared with
the model obtained using our methodology.

The last steps of our scheme is the a priori assessment of each model
using statistical methods (Procheck, Whatcheck, Verify 3D) and its final
validation by the comparison with the crystallographic structure.
Effective start/end date1/10/9930/09/03


  • alignment
  • homology modelling
  • Structure prediction
  • consensus
  • expert system