Detection of unrelated proteins in sequences multiple alignments by using predicted secondary structures

Motivation: Multiple sequence alignments are essential tools for establishing the homology relations between proteins. Essential amino acids for the function and/or the structure are generally conserved, thus providing key arguments to help in protein characterization. However for distant proteins, it is more difficult to establish, in a reliable way, the homology relations that may exist between them. In this article, we show that secondary structure prediction is a valuable way to validate protein families at low identity rate. Results: We show that the analysis of the secondary structures compatibility is a reliable way to discard non-related proteins in low identity multiple alignment.