While predicting 3D structure of a protein through homology modeling, the most important step is Multiple sequence Alignment of template sequences with target protein(whose 3D model is to be predicted). At this step according to my book for Multiple sequence Alignment, T-coffee or Praline soft-wares are used which are heuristic in nature and based on progressive alignment method.As we know that their are always concerns regarding sensitivity and specificity of heuristic type of Algorithms.So my query is that at this alignment step which is very critical in 3D modelling can we use Exhaustive algorithms to yield good models.
Exhaustive multiple sequence alignment is not feasible, at least with dynamic programming. All programs use heuristics. Some allow you to include structural information like helices and beta strands.
If you want to better align your target and template (and they are not very similar), make a multiple sequence alignment including other homologs. The target-template alignment will be more reliable.