Treelet kernel incorporating cyclic, stereo and inter pattern information in chemoinformatics

Abstract

Chemoinformatics is a research field concerned with the study of physical or biological molecular properties through computer science's research fields such as machine learning and graph theory. From this point of view, graph kernels provide a nice framework which allows to naturally combine machine learning and graph theory techniques. Graph kernels based on bags of patterns have proven their efficiency on several problems both in terms of accuracy and computational time. Treelet kernel is a graph kernel based on a bag of small subtrees. We propose in this paper several extensions of this kernel devoted to chemoinformatics problems. These extensions aim to weight each pattern according to its influence, to include the comparison of non-isomorphic patterns, to include stereo information and finally to explicitly encode cyclic information into kernel computation.

Approximate Graph Edit Distance Computation Combining Bipartite Matching and Exact Neighborhood Substructure Distance

“…However, a drawback of this approach is that graphs encode complex objects using nodes and relationships between nodes and a large amount of information is lost when graphs are transformed in vectors. Hence, instead of defining an explicit embedding of graphs, an alternative approach consists in using graph kernels [4] which correspond to a scalar product between two implicit embedding of graphs. Any graph kernel, which can be seen as a similarity measure between graphs, can then be used in any machine learning methods which access to data only through scalar products.…”

Ordered Decompositional DAG kernels enhancements

“…Complexity RW [34] O(|V | 3 ) SP [37] O(|V | 4 ) WL-SP [38] O(|V | 4 ) 3-Graphlet [40] O(|V | 3 ) Treelet [41] O(|V |ρ 5 ) FS [24,38] O Table 1: Computational complexity of the Shortest Path, the 3-Graphlet, the fast Subtree, the NSPDK, the ODD ST and ODD ST + kernels. *: considering ρ constant; **: with high constants.…”

“…The complexity of the kernel is O(|V 1 ||V 2 |hρ 2ρ ), where h is the depth of the visit. Finally, [41] proposed the treelet kernel, based on frequent pattern mining of tree-substructures. The kernel implementation considers subtrees with a maximum of 6 nodes, and its computational complexity is O(nρ 5 ).…”

Predicting activity approach based on new atoms similarity kernel function

“…Several extensions of Treelet kernel (Gaüzère et al, 2012) were proposed by (Gaüzère et al, 2015) which can be used to solve chemoinformatics problems. These extensions aim to weight each pattern according to its influence, to include the comparison of non-isomorphic patterns, to include stereo information and finally to explicitly encode cyclic information into kernel computation.…”