David Hoksza and Jakub Galgonek
View Full Paper
Protein structure, similarity, SCOP classiﬁcation, TM-score
Finding similarity between a pair of protein structures is one of the fundamental tasks in many areas of bioinformatical research such as protein structure prediction, function mapping, etc. We propose a method for ﬁnding pairing of amino acids based on densities of the structures and we also propose a modiﬁcation to the original template modeling-score (TM-Score) rotation algorithm that assess similarity score to this alignment. Proposed modiﬁcation is faster than TM and comparably robust according to non-optimal parts in the alignment. We measure the qualities of the algorithm in terms of structural classiﬁcation of proteins (SCOP) classiﬁcation accuracy. Regarding the accuracy, our solution outperforms the contemporary solutions at two out of three tested levels of the SCOP hierarchy.