Dear Anh,

I strongly encourage you to study classical papers describing these methods. All of them are using some kind of algorithmic solutions (similarity measures) However, what your professor expresses are traditional methods of building human knowledge, and in essence, his personal habits or preferences. Many researchers struggled with your question and some consensus was attained. I personally find this testament as baseless or rather without sufficient base. Why I have such views. When we ask a question we must ask about the assumptions we invested in formulation of the problem. In measuring the accuracy of the particular method there was a very reasonable assumption used: similar proteins are similar in many dimensions/characteristics such as activity, biological use, amino-acid content, folding, etc. Unfortunately every assumption has its own limitations. Generally these different characteristics are collinear which means have a strong correlation ( in another words they are useful). However, in studying problems of activity, or protein folding it turned out that very small changes have sometimes profound results. For instance, look up the history of a Paracelsus challenge. Small number of mutations change the fold completely. I even myself found and described the sequences that are ambivalent to folding which means they do not correlate with predictions. The same applies to the uniqueness of function. It happens that the same fold have different activities, and vastly different folds have similar or identical function. There are multifunctional enzymes processing many substrates and additionally regulated by effectors changing their activity. So, in conclusion it may or may not be applicable whether one or other method is superior as with increase evolutionary distance there would be a grey area or similarity, dissimilarity. Try to find papers by Adam Godzik and remote homologies.

Good Luck.

Bog

I wonder why did not you ask the same question to the same professor?