Algorithms for Regular Tree Grammar Network Search and Their Application to Mining Human–viral Infection Patterns

Abstract

Network querying is a powerful approach to mine molecular interaction networks. Most state-of-the-art network querying tools either confine the search to a prespecified topology in the form of some template subnetwork, or do not specify any topological constraints at all. Another approach is grammar-based queries, which are more flexible and expressive as they allow for expressing the topology of the sought pattern according to some grammar-based logic. Previous grammar-based network querying tools were confined to the identification of paths. In this article, we extend the patterns identified by grammar-based query approaches from paths to trees. For this, we adopt a higher order query descriptor in the form of a regular tree grammar (RTG). We introduce a novel problem and propose an algorithm to search a given graph for the k highest scoring subgraphs matching a tree accepted by an RTG. Our algorithm is based on the combination of dynamic programming with color coding, and includes an extension of previous k-best parsing optimization approaches to avoid isomorphic trees in the output. We implement the new algorithm and exemplify its application to mining viral infection patterns within molecular interaction networks. Our code is available online.

ResponseNet v.3: revealing signaling and regulatory pathways connecting your proteins and genes across human tissues

“…The two sets can vary in size from few to hundreds. Examples include analysis of disease-causing mutations and genes that are differentially expressed in patients (18); the proteins affected by a virus and the host response to viral infection (32,33); or the results of a genetic screen and the cellular response to perturbation (15). The user then selects the interactome to be searched, or uploads a weighted interactome.…”

A Survey on Syntactic Pattern Recognition Methods in Bioinformatics

“…Tree grammars were applied for the modeling of multiple biomolecular structures in [210,211], for the computation of exact RNA shape probabilities in [93], for RNA analysis [129] and for RNA pseudoknot comparison in [152]. The mining of human-viral infection patterns was performed with the help of regular tree grammars in [182]. Rectangle tree grammars were used for predicting RNA secondary structures in [127].…”