DeepNF

Deep network fusion for protein function prediction

Vladimir Gligorijević, Meet Barot, Richard Bonneau

Research output: Contribution to journalArticle

Abstract

Motivation The prevalence of high-throughput experimental methods has resulted in an abundance of large-scale molecular and functional interaction networks. The connectivity of these networks provides a rich source of information for inferring functional annotations for genes and proteins. An important challenge has been to develop methods for combining these heterogeneous networks to extract useful protein feature representations for function prediction. Most of the existing approaches for network integration use shallow models that encounter difficulty in capturing complex and highly non-linear network structures. Thus, we propose deepNF, a network fusion method based on Multimodal Deep Autoencoders to extract high-level features of proteins from multiple heterogeneous interaction networks. Results We apply this method to combine STRING networks to construct a common low-dimensional representation containing high-level protein features. We use separate layers for different network types in the early stages of the multimodal autoencoder, later connecting all the layers into a single bottleneck layer from which we extract features to predict protein function. We compare the cross-validation and temporal holdout predictive performance of our method with state-of-the-art methods, including the recently proposed method Mashup. Our results show that our method outperforms previous methods for both human and yeast STRING networks. We also show substantial improvement in the performance of our method in predicting gene ontology terms of varying type and specificity. Availability and implementation deepNF is freely available at: https://github.com/VGligorijevic/deepNF. Supplementary informationSupplementary dataare available at Bioinformatics online.

Original languageEnglish (US)
Pages (from-to)3873-3881
Number of pages9
JournalBioinformatics
Volume34
Issue number22
DOIs
StatePublished - Nov 15 2018

Fingerprint

Fusion
Fusion reactions
Proteins
Protein
Prediction
Genes
Nonlinear networks
Heterogeneous networks
Complex networks
Bioinformatics
Yeast
Ontology
Throughput
Availability
Gene Ontology
Heterogeneous Networks
Computational Biology
Interaction
Network Structure
Cross-validation

ASJC Scopus subject areas

  • Statistics and Probability
  • Biochemistry
  • Molecular Biology
  • Computer Science Applications
  • Computational Theory and Mathematics
  • Computational Mathematics

Cite this

DeepNF : Deep network fusion for protein function prediction. / Gligorijević, Vladimir; Barot, Meet; Bonneau, Richard.

In: Bioinformatics, Vol. 34, No. 22, 15.11.2018, p. 3873-3881.

Research output: Contribution to journalArticle

Gligorijević, Vladimir ; Barot, Meet ; Bonneau, Richard. / DeepNF : Deep network fusion for protein function prediction. In: Bioinformatics. 2018 ; Vol. 34, No. 22. pp. 3873-3881.
@article{fb269b82155b4358bfafb7ee3cbf65c4,
title = "DeepNF: Deep network fusion for protein function prediction",
abstract = "Motivation The prevalence of high-throughput experimental methods has resulted in an abundance of large-scale molecular and functional interaction networks. The connectivity of these networks provides a rich source of information for inferring functional annotations for genes and proteins. An important challenge has been to develop methods for combining these heterogeneous networks to extract useful protein feature representations for function prediction. Most of the existing approaches for network integration use shallow models that encounter difficulty in capturing complex and highly non-linear network structures. Thus, we propose deepNF, a network fusion method based on Multimodal Deep Autoencoders to extract high-level features of proteins from multiple heterogeneous interaction networks. Results We apply this method to combine STRING networks to construct a common low-dimensional representation containing high-level protein features. We use separate layers for different network types in the early stages of the multimodal autoencoder, later connecting all the layers into a single bottleneck layer from which we extract features to predict protein function. We compare the cross-validation and temporal holdout predictive performance of our method with state-of-the-art methods, including the recently proposed method Mashup. Our results show that our method outperforms previous methods for both human and yeast STRING networks. We also show substantial improvement in the performance of our method in predicting gene ontology terms of varying type and specificity. Availability and implementation deepNF is freely available at: https://github.com/VGligorijevic/deepNF. Supplementary informationSupplementary dataare available at Bioinformatics online.",
author = "Vladimir Gligorijević and Meet Barot and Richard Bonneau",
year = "2018",
month = "11",
day = "15",
doi = "10.1093/bioinformatics/bty440",
language = "English (US)",
volume = "34",
pages = "3873--3881",
journal = "Bioinformatics",
issn = "1367-4803",
publisher = "Oxford University Press",
number = "22",

}

TY - JOUR

T1 - DeepNF

T2 - Deep network fusion for protein function prediction

AU - Gligorijević, Vladimir

AU - Barot, Meet

AU - Bonneau, Richard

PY - 2018/11/15

Y1 - 2018/11/15

N2 - Motivation The prevalence of high-throughput experimental methods has resulted in an abundance of large-scale molecular and functional interaction networks. The connectivity of these networks provides a rich source of information for inferring functional annotations for genes and proteins. An important challenge has been to develop methods for combining these heterogeneous networks to extract useful protein feature representations for function prediction. Most of the existing approaches for network integration use shallow models that encounter difficulty in capturing complex and highly non-linear network structures. Thus, we propose deepNF, a network fusion method based on Multimodal Deep Autoencoders to extract high-level features of proteins from multiple heterogeneous interaction networks. Results We apply this method to combine STRING networks to construct a common low-dimensional representation containing high-level protein features. We use separate layers for different network types in the early stages of the multimodal autoencoder, later connecting all the layers into a single bottleneck layer from which we extract features to predict protein function. We compare the cross-validation and temporal holdout predictive performance of our method with state-of-the-art methods, including the recently proposed method Mashup. Our results show that our method outperforms previous methods for both human and yeast STRING networks. We also show substantial improvement in the performance of our method in predicting gene ontology terms of varying type and specificity. Availability and implementation deepNF is freely available at: https://github.com/VGligorijevic/deepNF. Supplementary informationSupplementary dataare available at Bioinformatics online.

AB - Motivation The prevalence of high-throughput experimental methods has resulted in an abundance of large-scale molecular and functional interaction networks. The connectivity of these networks provides a rich source of information for inferring functional annotations for genes and proteins. An important challenge has been to develop methods for combining these heterogeneous networks to extract useful protein feature representations for function prediction. Most of the existing approaches for network integration use shallow models that encounter difficulty in capturing complex and highly non-linear network structures. Thus, we propose deepNF, a network fusion method based on Multimodal Deep Autoencoders to extract high-level features of proteins from multiple heterogeneous interaction networks. Results We apply this method to combine STRING networks to construct a common low-dimensional representation containing high-level protein features. We use separate layers for different network types in the early stages of the multimodal autoencoder, later connecting all the layers into a single bottleneck layer from which we extract features to predict protein function. We compare the cross-validation and temporal holdout predictive performance of our method with state-of-the-art methods, including the recently proposed method Mashup. Our results show that our method outperforms previous methods for both human and yeast STRING networks. We also show substantial improvement in the performance of our method in predicting gene ontology terms of varying type and specificity. Availability and implementation deepNF is freely available at: https://github.com/VGligorijevic/deepNF. Supplementary informationSupplementary dataare available at Bioinformatics online.

UR - http://www.scopus.com/inward/record.url?scp=85056389955&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056389955&partnerID=8YFLogxK

U2 - 10.1093/bioinformatics/bty440

DO - 10.1093/bioinformatics/bty440

M3 - Article

VL - 34

SP - 3873

EP - 3881

JO - Bioinformatics

JF - Bioinformatics

SN - 1367-4803

IS - 22

ER -