All-paths graph kernel for protein-protein interaction extraction with evaluation of cross-corpus learning

By Antti Airola, Sampo Pyysalo, Jari Björne, Tapio Pahikkala, Filip Ginter, Tapio Salakoski. BMC Bioinformatics, 2008

313 citations recorded by the source corpus.

BackgroundAutomated extraction of protein-protein interactions (PPI) is an important and widely studied task in biomedical text mining. We propose a graph kernel based approach for this task. In contrast to earlier approaches to PPI extraction, the introduced all-paths graph kernel has the capability to make use of full, general dependency graphs representing the sentence structure.ResultsWe evaluate the proposed method on five publicly available PPI corpora, providing the most comprehensive evaluation done for a machine learning based PPI-extraction system. We additionally perform a detailed evaluation of the effects of training and testing on different resources, providing insight into the challenges involved in applying a system beyond the data it was trained on. Our method is shown to achieve state-of-the-art performance with respect to comparable evaluations, with 56.4 F-score and 84.8 AUC on the AImed corpus.ConclusionWe show that the graph kernel approach performs on state-of-the-art level in PPI extraction, and note the possible extension to the task of extracting complex interactions. Cross-corpus results provide further insight into how the learning generalizes beyond individual corpora. Further, we identify several pitfalls that can make evaluations of PPI-extraction systems incomparable, or even invalid. These include incorrect cross-validation strategies and problems related to comparing F-score results achieved on different evaluation resources. Recommendations for avoiding these pitfalls are provided.

Subfields: Molecular Biology

View Paper: https://www.semanticscholar.org/paper/987bc971e50b893cc3e0a6698b44d0c3baa28cc8

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

1. Zhiyong Lu
2. Yuan Luo
3. Ozlem Uzuner
4. Peter Szolovits
5. Chih-Hsuan Wei
6. Yifan Peng
7. Lei Hua
8. Chanqin Quan

Citations in the corpus

• Bridging semantics and syntax with graph algorithms - state-of-the-art of extracting biomedical relations (2017, methodology): “Abbreviation used in this table include: CoRef, coreference resolution; CRF, conditional random field; HMM, hidden Markov model; APG, all paths graph kernel [20];”
• Deep learning for extracting protein-protein interactions from biomedical literature (2017, methodology): “Rows 3 and 4 report the results of two previous best single kernel-based methods, an APG kernel (Airola et al., 2008; Tikk et al., 2010) and an edit kernel (Peng et al., 2015).”
• Improving chemical disease relation extraction with rich features and weakly labeled data (2016, methodology): “Most useful feature sets include lexical information and various linguistic/semantic parser outputs [1, 2, 15, 23, 38].”
• Deep Learning with Minimal Training Data: TurkuNLP Entry in the BioNLP Shared Task 2016 (2016, methodology): “This observation has served as the basis for many successful relation extraction approaches in both general and biomedical domain NLP (Bunescu and Mooney, 2005; Airola et al., 2008; Nguyen et al., 2009; Chowdhury et al., 2011).”
• SemEval-2013 Task 9 : Extraction of Drug-Drug Interactions from Biomedical Texts (DDIExtraction 2013) (2013, methodology): “Its system combines several kernel methods (APG (Airola et al., 2008) and Shallow Linguistic Kernel (SL) (Giuliano et al., 2006) among others), the Turku Event Extraction system (TEES) (Bj¨orne et al., 2011) 8 and the Moara system (Neves et al., 2009).”
• A Comprehensive Benchmark of Kernel Methods to Extract Protein–Protein Interactions from Literature (2010, methodology): “Recently, some authors criticized the F-score as performance measure, because it is very sensitive to the ratio of positive/ negative pairs in the corpus [17,36], and it is less stable to parameter modifications than AUC.”
• Bridging semantics and syntax with graph algorithms —state-of-the-art of extracting biomedical relations (2015, methodology): “To keep up with the exponential growth of the literature, automated methods have been applied to mining protein-protein interactions [15,16], gene-phenotype associations [17,18], gene ontology [19], and pathway information [20], which we collectively call biomolecular information extraction.”
• A Shortest Dependency Path Based Convolutional Neural Network for Protein-Protein Relation Extraction (2016, methodology): “The kernels include bag-of-words kernel [10], all-path kernel [11], subset-tree kernel [12], edit-distance kernel [8], and graph kernel [13], and they have shown effectiveness in PPI task.”