↓ Skip to main content

Neural networks for link prediction in realistic biomedical graphs: a multi-dimensional evaluation of graph embedding-based approaches

Overview of attention for article published in BMC Bioinformatics, May 2018
Altmetric Badge

About this Attention Score

  • Average Attention Score compared to outputs of the same age
  • Average Attention Score compared to outputs of the same age and source

Mentioned by

twitter
2 tweeters

Citations

dimensions_citation
23 Dimensions

Readers on

mendeley
66 Mendeley
citeulike
1 CiteULike
You are seeing a free-to-access but limited selection of the activity Altmetric has collected about this research output. Click here to find out more.
Title
Neural networks for link prediction in realistic biomedical graphs: a multi-dimensional evaluation of graph embedding-based approaches
Published in
BMC Bioinformatics, May 2018
DOI 10.1186/s12859-018-2163-9
Pubmed ID
Authors

Gamal Crichton, Yufan Guo, Sampo Pyysalo, Anna Korhonen

Abstract

Link prediction in biomedical graphs has several important applications including predicting Drug-Target Interactions (DTI), Protein-Protein Interaction (PPI) prediction and Literature-Based Discovery (LBD). It can be done using a classifier to output the probability of link formation between nodes. Recently several works have used neural networks to create node representations which allow rich inputs to neural classifiers. Preliminary works were done on this and report promising results. However they did not use realistic settings like time-slicing, evaluate performances with comprehensive metrics or explain when or why neural network methods outperform. We investigated how inputs from four node representation algorithms affect performance of a neural link predictor on random- and time-sliced biomedical graphs of real-world sizes (∼ 6 million edges) containing information relevant to DTI, PPI and LBD. We compared the performance of the neural link predictor to those of established baselines and report performance across five metrics. In random- and time-sliced experiments when the neural network methods were able to learn good node representations and there was a negligible amount of disconnected nodes, those approaches outperformed the baselines. In the smallest graph (∼ 15,000 edges) and in larger graphs with approximately 14% disconnected nodes, baselines such as Common Neighbours proved a justifiable choice for link prediction. At low recall levels (∼ 0.3) the approaches were mostly equal, but at higher recall levels across all nodes and average performance at individual nodes, neural network approaches were superior. Analysis showed that neural network methods performed well on links between nodes with no previous common neighbours; potentially the most interesting links. Additionally, while neural network methods benefit from large amounts of data, they require considerable amounts of computational resources to utilise them. Our results indicate that when there is enough data for the neural network methods to use and there are a negligible amount of disconnected nodes, those approaches outperform the baselines. At low recall levels the approaches are mostly equal but at higher recall levels and average performance at individual nodes, neural network approaches are superior. Performance at nodes without common neighbours which indicate more unexpected and perhaps more useful links account for this.

Twitter Demographics

The data shown below were collected from the profiles of 2 tweeters who shared this research output. Click here to find out more about how the information was compiled.

Mendeley readers

The data shown below were compiled from readership statistics for 66 Mendeley readers of this research output. Click here to see the associated Mendeley record.

Geographical breakdown

Country Count As %
Unknown 66 100%

Demographic breakdown

Readers by professional status Count As %
Student > Master 16 24%
Student > Ph. D. Student 14 21%
Student > Bachelor 8 12%
Researcher 7 11%
Student > Doctoral Student 4 6%
Other 6 9%
Unknown 11 17%
Readers by discipline Count As %
Computer Science 33 50%
Biochemistry, Genetics and Molecular Biology 6 9%
Agricultural and Biological Sciences 3 5%
Engineering 2 3%
Chemistry 2 3%
Other 7 11%
Unknown 13 20%

Attention Score in Context

This research output has an Altmetric Attention Score of 2. This is our high-level measure of the quality and quantity of online attention that it has received. This Attention Score, as well as the ranking and number of research outputs shown below, was calculated when the research output was last mentioned on 28 May 2018.
All research outputs
#7,835,177
of 12,996,278 outputs
Outputs from BMC Bioinformatics
#3,118
of 4,847 outputs
Outputs of similar age
#150,667
of 271,049 outputs
Outputs of similar age from BMC Bioinformatics
#8
of 22 outputs
Altmetric has tracked 12,996,278 research outputs across all sources so far. This one is in the 37th percentile – i.e., 37% of other outputs scored the same or lower than it.
So far Altmetric has tracked 4,847 research outputs from this source. They receive a mean Attention Score of 4.8. This one is in the 31st percentile – i.e., 31% of its peers scored the same or lower than it.
Older research outputs will score higher simply because they've had more time to accumulate mentions. To account for age we can compare this Altmetric Attention Score to the 271,049 tracked outputs that were published within six weeks on either side of this one in any source. This one is in the 40th percentile – i.e., 40% of its contemporaries scored the same or lower than it.
We're also able to compare this research output to 22 others from the same source and published within six weeks on either side of this one. This one is in the 45th percentile – i.e., 45% of its contemporaries scored the same or lower than it.