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

Airola, Antti; Pyysalo, Sampo; Björne, Jari; Pahikkala, Tapio; Ginter, Filip; Salakoski, Tapio

doi:10.1186/1471-2105-9-s11-s2

Cited by 225 publications

(314 citation statements)

References 23 publications

(29 reference statements)

Supporting

Mentioning

308

Contrasting

Unclassified

Order By: Relevance

“…1000 positive relations and 4834 negative examples. The difference between the two reference sets make it impossible to compare our results in Table 7 to this state-of-the-art bRE system as claimed by [1] (with 56.4 F-score). Therefore we re-experiment ASK on this new AIMED relation corpus with both local and global ASK using the mismatch or spectrum kernel.…”

Section: Task 3: Ppi Extraction Relation-level: Aimedmentioning

confidence: 99%

“…Furthermore as pointed out by [1], though the AIMED corpus has been applied in numerous evaluations for PPI relation extraction, the datasets used in different papers varied largely due to diverse postprocessing rules used to create the relation-level examples. For instance, the corpus used to test our ASK in Table 7 was downloaded from [9] which contains 4026 examples with 951 as positive and 3075 as negatives.…”

Section: Task 3: Ppi Extraction Relation-level: Aimedmentioning

confidence: 99%

“…For instance, the corpus used to test our ASK in Table 7 was downloaded from [9] which contains 4026 examples with 951 as positive and 3075 as negatives. However, the AIMED corpus used in [1] includes more relation examples, i.e. 1000 positive relations and 4834 negative examples.…”

Section: Task 3: Ppi Extraction Relation-level: Aimedmentioning

confidence: 99%

“…For bRE tasks at the relation level, most systems in the literature are rulebased, cooccurrence-based or hybrid approaches (survey in [29]). Recently several researchers proposed the all-paths graph kernel [1], or an ensemble of multiple kernels and parsers [21], which were reported to yield good results.…”

Section: Introductionmentioning

confidence: 99%

“…In order to identify PPI events, the tasks aim to: (1) retrieve PubMed abstracts describing PPIs; (2) classify text sentences as PPI relevant or not relevant; (3) when protein entities have been recognized in the sentence, extract which protein-protein pairs having interaction relationship, i.e. pairwise PPI relations from the sentence.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Semi-supervised Abstraction-Augmented String Kernel for Multi-level Bio-Relation Extraction

Kuksa

Bai

et al. 2010

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Bio-relation extraction (bRE), an important goal in bio-text mining, involves subtasks identifying relationships between bio-entities in text at multiple levels, e.g., at the article, sentence or relation level. A key limitation of current bRE systems is that they are restricted by the availability of annotated corpora. In this work we introduce a semisupervised approach that can tackle multi-level bRE via string comparisons with mismatches in the string kernel framework. Our string kernel implements an abstraction step, which groups similar words to generate more abstract entities, which can be learnt with unlabeled data. Specifically, two unsupervised models are proposed to capture contextual (local or global) semantic similarities between words from a large unannotated corpus. This Abstraction-augmented String Kernel (ASK) allows for better generalization of patterns learned from annotated data and provides a unified framework for solving bRE with multiple degrees of detail. ASK shows effective improvements over classic string kernels on four datasets and achieves state-of-the-art bRE performance without the need for complex linguistic features.

show abstract

Section: Task 3: Ppi Extraction Relation-level: Aimedmentioning

confidence: 99%

Section: Task 3: Ppi Extraction Relation-level: Aimedmentioning

confidence: 99%

Section: Task 3: Ppi Extraction Relation-level: Aimedmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Semi-supervised Abstraction-Augmented String Kernel for Multi-level Bio-Relation Extraction

Kuksa

Bai

et al. 2010

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

Graph Kernels for Molecular Similarity

Rupp

Schneider

2010

Molecular Informatics

View full text Add to dashboard Cite

Molecular similarity measures are important for many cheminformatics applications like ligand-based virtual screening and quantitative structure-property relationships. Graph kernels are formal similarity measures defined directly on graphs, such as the (annotated) molecular structure graph. Graph kernels are positive semi-definite functions, i.e., they correspond to inner products. This property makes them suitable for use with kernel-based machine learning algorithms such as support vector machines and Gaussian processes. We review the major types of kernels between graphs (based on random walks, subgraphs, and optimal assignments, respectively), and discuss their advantages, limitations, and successful applications in cheminformatics.

show abstract

Ranking support vector machine for multiple kernels output combination in protein–protein interaction extraction from biomedical literature

Yang

Lin

et al. 2011

Proteomics

View full text Add to dashboard Cite

Increased climatic variability is resulting in an increase of both the frequency and the magnitude of extreme climate events. Therefore, cereals may be exposed to more than one stress event in the growing season, which may ultimately affect crop yield and quality. Here, effects are reported of interaction of water deficits and/or a high-temperature event (32°C) during vegetative growth (terminal spikelet) with either of these stress events applied during generative growth (anthesis) in wheat. Influence of combinations of stress on protein fractions (albumins, globulins, gliadins and glutenins) in grains and stress-induced changes on the albumin and gliadin proteomes were investigated by 2-DE and MS. The synthesis of individual protein fractions was shown to be affected by both the type and time of the applied stresses. Identified drought or high-temperature-responsive proteins included proteins involved in primary metabolism, storage and stress response such as late embryogenesis abundant proteins, peroxiredoxins and α-amylase/trypsin inhibitors. Several proteins, e.g. heat shock protein and 14-3-3 protein changed in abundance only under multiple high temperatures.

show abstract

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

Cited by 225 publications

References 23 publications

Semi-supervised Abstraction-Augmented String Kernel for Multi-level Bio-Relation Extraction

Semi-supervised Abstraction-Augmented String Kernel for Multi-level Bio-Relation Extraction

Graph Kernels for Molecular Similarity

Ranking support vector machine for multiple kernels output combination in protein–protein interaction extraction from biomedical literature

Contact Info

Product

Resources

About