Comparing two sequences is generally based on their local score and the corresponding p-value. Normally, the local score is obtain from an alignment that is not started near the end of either sequence. So, it is necessary to consider this fact in p-value calculation. The edge effect correction (EEC) (finite size correction) is an appropriate way to improve the significance of sequences alignment. In this paper, the EEC is applied to a method of evaluating sequence similarity, the h-tuple method, then, the results of the corrected h-tuple method are compared with the ones based on the extreme value theory in a real database. The receiver operating characteristic (ROC) curve reveals that the corrected h-tuple method is more accurate.
