BTNET: Boosted tree based gene regulatory network inference algorithm using time-course measurement data

Sungjoon Park, Jung Min Kim, Wonho Shin, Sung Won Han, Minji Jeon, Hyun Jin Jang, Ik Soon Jang, Jaewoo Kang

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Background: Identifying gene regulatory networks is an important task for understanding biological systems. Time-course measurement data became a valuable resource for inferring gene regulatory networks. Various methods have been presented for reconstructing the networks from time-course measurement data. However, existing methods have been validated on only a limited number of benchmark datasets, and rarely verified on real biological systems. Results: We first integrated benchmark time-course gene expression datasets from previous studies and reassessed the baseline methods. We observed that GENIE3-time, a tree-based ensemble method, achieved the best performance among the baselines. In this study, we introduce BTNET, a boosted tree based gene regulatory network inference algorithm which improves the state-of-the-art. We quantitatively validated BTNET on the integrated benchmark dataset. The AUROC and AUPR scores of BTNET were higher than those of the baselines. We also qualitatively validated the results of BTNET through an experiment on neuroblastoma cells treated with an antidepressant. The inferred regulatory network from BTNET showed that brachyury, a transcription factor, was regulated by fluoxetine, an antidepressant, which was verified by the expression of its downstream genes. Conclusions: We present BTENT that infers a GRN from time-course measurement data using boosting algorithms. Our model achieved the highest AUROC and AUPR scores on the integrated benchmark dataset. We further validated BTNET qualitatively through a wet-lab experiment and showed that BTNET can produce biologically meaningful results.

Original languageEnglish
Article number20
JournalBMC Systems Biology
Volume12
DOIs
Publication statusPublished - 2018 Mar 19

Fingerprint

Gene Regulatory Networks
Gene Regulatory Network
Benchmarking
Time measurement
Genes
Benchmark
Biological systems
Baseline
Biological Systems
Antidepressive Agents
Transcription factors
Ensemble Methods
Gene expression
Fluoxetine
Regulatory Networks
Boosting
Transcription Factor
Neuroblastoma
Experiments
Gene Expression

Keywords

  • Boosted tree
  • Gene regulatory network inference
  • Time course

ASJC Scopus subject areas

  • Structural Biology
  • Modelling and Simulation
  • Molecular Biology
  • Computer Science Applications
  • Applied Mathematics

Cite this

BTNET : Boosted tree based gene regulatory network inference algorithm using time-course measurement data. / Park, Sungjoon; Kim, Jung Min; Shin, Wonho; Han, Sung Won; Jeon, Minji; Jang, Hyun Jin; Jang, Ik Soon; Kang, Jaewoo.

In: BMC Systems Biology, Vol. 12, 20, 19.03.2018.

Research output: Contribution to journalArticle

Park, Sungjoon ; Kim, Jung Min ; Shin, Wonho ; Han, Sung Won ; Jeon, Minji ; Jang, Hyun Jin ; Jang, Ik Soon ; Kang, Jaewoo. / BTNET : Boosted tree based gene regulatory network inference algorithm using time-course measurement data. In: BMC Systems Biology. 2018 ; Vol. 12.
@article{9bfff7723caf465ea8bf9773e90888e2,
title = "BTNET: Boosted tree based gene regulatory network inference algorithm using time-course measurement data",
abstract = "Background: Identifying gene regulatory networks is an important task for understanding biological systems. Time-course measurement data became a valuable resource for inferring gene regulatory networks. Various methods have been presented for reconstructing the networks from time-course measurement data. However, existing methods have been validated on only a limited number of benchmark datasets, and rarely verified on real biological systems. Results: We first integrated benchmark time-course gene expression datasets from previous studies and reassessed the baseline methods. We observed that GENIE3-time, a tree-based ensemble method, achieved the best performance among the baselines. In this study, we introduce BTNET, a boosted tree based gene regulatory network inference algorithm which improves the state-of-the-art. We quantitatively validated BTNET on the integrated benchmark dataset. The AUROC and AUPR scores of BTNET were higher than those of the baselines. We also qualitatively validated the results of BTNET through an experiment on neuroblastoma cells treated with an antidepressant. The inferred regulatory network from BTNET showed that brachyury, a transcription factor, was regulated by fluoxetine, an antidepressant, which was verified by the expression of its downstream genes. Conclusions: We present BTENT that infers a GRN from time-course measurement data using boosting algorithms. Our model achieved the highest AUROC and AUPR scores on the integrated benchmark dataset. We further validated BTNET qualitatively through a wet-lab experiment and showed that BTNET can produce biologically meaningful results.",
keywords = "Boosted tree, Gene regulatory network inference, Time course",
author = "Sungjoon Park and Kim, {Jung Min} and Wonho Shin and Han, {Sung Won} and Minji Jeon and Jang, {Hyun Jin} and Jang, {Ik Soon} and Jaewoo Kang",
year = "2018",
month = "3",
day = "19",
doi = "10.1186/s12918-018-0547-0",
language = "English",
volume = "12",
journal = "BMC Systems Biology",
issn = "1752-0509",
publisher = "BioMed Central",

}

TY - JOUR

T1 - BTNET

T2 - Boosted tree based gene regulatory network inference algorithm using time-course measurement data

AU - Park, Sungjoon

AU - Kim, Jung Min

AU - Shin, Wonho

AU - Han, Sung Won

AU - Jeon, Minji

AU - Jang, Hyun Jin

AU - Jang, Ik Soon

AU - Kang, Jaewoo

PY - 2018/3/19

Y1 - 2018/3/19

N2 - Background: Identifying gene regulatory networks is an important task for understanding biological systems. Time-course measurement data became a valuable resource for inferring gene regulatory networks. Various methods have been presented for reconstructing the networks from time-course measurement data. However, existing methods have been validated on only a limited number of benchmark datasets, and rarely verified on real biological systems. Results: We first integrated benchmark time-course gene expression datasets from previous studies and reassessed the baseline methods. We observed that GENIE3-time, a tree-based ensemble method, achieved the best performance among the baselines. In this study, we introduce BTNET, a boosted tree based gene regulatory network inference algorithm which improves the state-of-the-art. We quantitatively validated BTNET on the integrated benchmark dataset. The AUROC and AUPR scores of BTNET were higher than those of the baselines. We also qualitatively validated the results of BTNET through an experiment on neuroblastoma cells treated with an antidepressant. The inferred regulatory network from BTNET showed that brachyury, a transcription factor, was regulated by fluoxetine, an antidepressant, which was verified by the expression of its downstream genes. Conclusions: We present BTENT that infers a GRN from time-course measurement data using boosting algorithms. Our model achieved the highest AUROC and AUPR scores on the integrated benchmark dataset. We further validated BTNET qualitatively through a wet-lab experiment and showed that BTNET can produce biologically meaningful results.

AB - Background: Identifying gene regulatory networks is an important task for understanding biological systems. Time-course measurement data became a valuable resource for inferring gene regulatory networks. Various methods have been presented for reconstructing the networks from time-course measurement data. However, existing methods have been validated on only a limited number of benchmark datasets, and rarely verified on real biological systems. Results: We first integrated benchmark time-course gene expression datasets from previous studies and reassessed the baseline methods. We observed that GENIE3-time, a tree-based ensemble method, achieved the best performance among the baselines. In this study, we introduce BTNET, a boosted tree based gene regulatory network inference algorithm which improves the state-of-the-art. We quantitatively validated BTNET on the integrated benchmark dataset. The AUROC and AUPR scores of BTNET were higher than those of the baselines. We also qualitatively validated the results of BTNET through an experiment on neuroblastoma cells treated with an antidepressant. The inferred regulatory network from BTNET showed that brachyury, a transcription factor, was regulated by fluoxetine, an antidepressant, which was verified by the expression of its downstream genes. Conclusions: We present BTENT that infers a GRN from time-course measurement data using boosting algorithms. Our model achieved the highest AUROC and AUPR scores on the integrated benchmark dataset. We further validated BTNET qualitatively through a wet-lab experiment and showed that BTNET can produce biologically meaningful results.

KW - Boosted tree

KW - Gene regulatory network inference

KW - Time course

UR - http://www.scopus.com/inward/record.url?scp=85044224185&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044224185&partnerID=8YFLogxK

U2 - 10.1186/s12918-018-0547-0

DO - 10.1186/s12918-018-0547-0

M3 - Article

C2 - 29560827

AN - SCOPUS:85044224185

VL - 12

JO - BMC Systems Biology

JF - BMC Systems Biology

SN - 1752-0509

M1 - 20

ER -