Mining Patterns from Glycan Structures
Ichigaku Takigawa, Kosuke Hashimoto, M. Shiga, Minoru I. Kanehisa, Hiroshi Mamitsuka · 2010
Glycans can be directed trees in which nodes are monosaccharides and edges are linkages extending between monosaccharides in the direction from the amino acid-connecting monosaccharide. We present an efficient method for mining frequent and statistically significant subtrees from glycan trees. The two key points of the method are: (1) It can reduce the number of redundant subtrees obtained by usual frequent subtree mining techniques and (2) can keep significant subtrees only by removing frequent but insignificant subtrees, like those with only one or a few nodes. We confirmed the efficiency of the approach in various manners, including biological significance. Our approach would be useful for mining unknown conserved patterns in larger glycan datasets to be obtained by a high-throughput manner in the future.