Enhancement and Reusage of Biomedical Knowledge Graph Subsets

Abstract

Knowledge Graphs (KGs) such as Wikidata act as a hub of information from multiple domains and disciplines, and is crowdsourced by multiple stakeholders. The vast amount of available information makes it difficult for researchers to manage the entire KG, which is also continually being edited. It is necessary to develop tools that extract subsets for domains of interest. These subsets will help researchers to reduce costs and time, making data of interest more accessible. In the last two BioHackathons (BH20, BH21), we have created prototypes to extract subsets easily applicable to Wikidata, as well as to define a map of the different approaches used to tackle this problem. Building on those outcomes, we aim to enhance subsetting in both definitions using Entity schemas based on Shape Expressions (ShEx) and extraction algorithms, with a special focus on the biomedical domain. Our first aim is to develop complex subsetting patterns based on qualifiers and references for enhancing credibility of datasets. Our second aim is to establish a faster subsetting extraction platform applying new algorithms based on Apache Spark and new tools like a document-oriented DBMS platform.

Wikidata subsetting: Approaches, tools, and evaluation

“…The RDF serializer in WDumper and WDSub is the same; however, the WDSub filtering system (based on ShEx) can parse quite complex 25 https://github.com/kg-subsetting/paper-wikidata-subsetting-2023/blob/dc2869c/performance-experiments/count_instances_tsv.py -accessed 10 June 2023. 26 filters at the SPARQL level, which creates a massive overhead. WDumper also has a better level of multithreading than WDSub.…”

“…SPARQL endpoints are usually slow and have run-time restrictions. Moreover, recursive data models are not supported in standard SPARQL implementations [20].…”

“…Subsetting is a recent research problem in KGs. To the best of our knowledge, the early demand for creating a biomedical subset of Wikidata was in 2017 [29], the subsetting discussions in the Wikidata biomedical community were concretely started at the 12th international SWAT4HCLS conference in 2019 by Andra Waagmeester et al [42] and then followed in Project 35 11 of ELIXIR BioHackathon-Europe 2020 [27], Project 21 12 of ELIXIR BioHackathon-Europe 2021, and Project 11 13 of ELIXIR BioHackathon-Europe 2022 [26].…”

“…A Python script 20 runs each tool three times separately from the moment of starting with the raw input dump to the moment it saves the output on disk. In this way, the time required for any indexing and pre-processing of the dump (if any), as well as the time of writing the output, is included in the extraction time, which is in line with the local and independent subsetting scope.…”

“…For KGTK which produces TSV outputs, a Python script (using pandas package) has been used . 25 For counting the number of instances and statements in the WDF output and the input dump (which are JSON files), a parallelized Python script 26 with efficient time consumption has been used. Note that while each Wikidata JSON dump has an RDF pair dump, these two different serializations are not identical [34].…”