I am quite confused by which one should be used in inferring phylogenetic relationship based on phylogenetic tree, consensus tree or the best tree. I have seen different people may used different strategies. Could anyone let me know which is better (or better in some situations)? Thanks so much!
They both contain relevant information.
The optimal topology is the tree that fits best with your data, the equivalent of a point-estimate form a statistical procedure
A consensus tree can gives you information on which relationships within your optimal tree are well supported (if you have used bootstraps to get your supports the values tell you how consistently each group is supported by your data, in a Bayesian framework the values are how probable the relationship is).
In practice, you don't want to be placing great weight on any interpretations of you data that hang on weakly supported relationships, so it's crucial you consider the support values on your tree (probably via a consensus tree, but perhaps by projecting bootstrap values on the optimal topology).