Optimizations for Computing Relatedness in Biomedical Heterogeneous Information Networks: SemNet 2.0

Abstract

Literature-based discovery (LBD) summarizes information and generates insight from large text corpuses. The SemNet framework utilizes a large heterogeneous information network or “knowledge graph” of nodes and edges to compute relatedness and rank concepts pertinent to a user-specified target. SemNet provides a way to perform multi-factorial and multi-scalar analysis of complex disease etiology and therapeutic identification using the 33+ million articles in PubMed. The present work improves the efficacy and efficiency of LBD for end users by augmenting SemNet to create SemNet 2.0. A custom Python data structure replaced reliance on Neo4j to improve knowledge graph query times by several orders of magnitude. Additionally, two randomized algorithms were built to optimize the HeteSim metric calculation for computing metapath similarity. The unsupervised learning algorithm for rank aggregation (ULARA), which ranks concepts with respect to the user-specified target, was reconstructed using derived mathematical proofs of correctness and probabilistic performance guarantees for optimization. The upgraded ULARA is generalizable to other rank aggregation problems outside of SemNet. In summary, SemNet 2.0 is a comprehensive open-source software for significantly faster, more effective, and user-friendly means of automated biomedical LBD. An example case is performed to rank relationships between Alzheimer’s disease and metabolic co-morbidities.

CompositeView: A Network-Based Visualization Tool

“…While neo4j is convenient for querying a database, it is known to be sluggish when performing multiple queries on a large graph. For example, the replacement of neo4j with nested Python libraries resulted in multiple orders of magnitude of speed up in the SemNet 2.0 biomedical knowledge graph analysis software [ 9 ]. Given that neo4j is mostly reliant on other embedded tools or external products, it does not have the built-in functionality to update composite scores on the fly as a user interacts with the data to examine varying levels of filtering or aggregation.…”

“…CompositeView was motivated by and originally developed for examining knowledge graph relationship ranking results, such as those produced by both SemNet version 1 [ 8 ] and SemNet version 2 [ 9 ]. As such, CompositeView enables SemNet users to gain deep insight from the results in a fraction of the time previously spent parsing through tables.…”

“…The Methods section details how CompositeView was constructed and how it is used, including a stress testing analysis for CompositeView’s data handling, processing, and efficiency. The Results section illustrates three distinct data set examples visualized with CompositeView: results from a SemNet [ 8 , 9 ] biomedical knowledge graph analysis of the two target nodes AD and hypothyroidism, Human Development Index (HDI) data [ 10 ], and cardiovascular disease (CVD) data [ 11 ]. Two of these three examples are completely removed from the study of networks, and the example involving HDI data is not strictly related to biomedical sciences.…”

“…SemNet is a software tool developed to conduct LBD on an expansive heterogeneous information network (knowledge graph) constructed from the Semantic Medline Database (SemMedDB) [ 8 , 9 ]. This database is, essentially, a table containing text-mined subject-predicate-object triples, extracted from article abstracts within PubMed, that convey some sort of relationship between biomedical concepts [ 17 ].…”

“…As a brief aside, because HeteSim is calculated on a per-metapath basis, and source and target nodes are often connected by multiple metapaths, the aggregation of HeteSim scores is necessary for a combined relevance value derived from multiple metapaths. This aggregation is typically the arithmetic mean of all HeteSim values; however, other aggregation methods (specifically randomized aggregation, in SemNet 2.0) have been used successfully [ 9 ].…”