Back

Biological Insights Knowledge Graph: an integrated knowledge graph to support drug development

Geleta, D.; Nikolov, A.; Edwards, G.; Gogleva, A.; Jackson, R.; Jansson, E.; Lamov, A.; Nilsson, S.; Pettersson, M.; Poroshin, V.; Rozemberczki, B.; Scrivener, T.; Ughetto, M.; Papa, E.

2021-11-01 systems biology
10.1101/2021.10.28.466262 bioRxiv
Show abstract

The use of knowledge graphs as a data source for machine learning methods to solve complex problems in life sciences has rapidly become popular in recent years. Our Biological Insights Knowledge Graph (BIKG) combines relevant data for drug development from public as well as internal data sources to provide insights for a range of tasks: from identifying new targets to repurposing existing drugs. Besides the common requirements to organisational knowledge graphs such as being able to capture the domain precisely and give the users the ability to search and query the data, the focus on handling multiple use cases and supporting use case-specific machine learning models presents additional challenges: the data models must also be streamlined for the performance of downstream tasks; graph content must be easily customisable for different use cases; different projections of the graph content are required to support a wider range of different consumption modes. In this paper we describe our main design choices in implementation of the BIKG graph and discuss different aspects of its life cycle: from graph construction to exploitation.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.