Multilabel classification of drug-like molecules via max-margin conditional random fields

Hongyu Su; Markus Heinonen; Juho Rousu

Multilabel classification of drug-like molecules via max-margin conditional random fields

Not published at Aalto University

University of Helsinki

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

Abstract

We present a multilabel learning approach for molecular classification, an important task in drug discovery. We use a conditional random field to model the dependencies between drug targets and discriminative training to separate correct multilabels from incorrect ones with a large margin. Efficient training of the model is ensured by conditional gradient optimization on the marginal dual polytope, using loopy belief propagation to find the steepest feasible ascent directions. In our experiments, the MMCRF method outperformed the support vector machine with state-of-the-art graph kernels on a dataset comprising of cancer inhibition potential of drug-like molecules against a large number cancer cell lines.

Original language	English
Title of host publication	Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010)
Subtitle of host publication	13-15 September, 2010, Helsinki, Finland
Editors	Petri Myllymäki, Teemu Roos, Tommi Jaakkola
Publisher	Helsinki Institute for Information Technology HIIT
Pages	265-272
ISBN (Electronic)	978-952-60-3314-3
Publication status	Published - 2010
MoE publication type	A4 Conference publication
Event	European Workshop on Probabilistic Graphical Models - Helsinki, Finland Duration: 13 Sept 2010 → 15 Sept 2010 Conference number: 5

Publication series

Name	HIIT Publications
Publisher	Helsinki Institute for Information Technology HIIT
Number	2
Volume	2010
ISSN (Electronic)	1458-946X

Workshop

Workshop	European Workshop on Probabilistic Graphical Models
Abbreviated title	PGM
Country/Territory	Finland
City	Helsinki
Period	13/09/2010 → 15/09/2010

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Cite this

Su, H., Heinonen, M., & Rousu, J. (2010). Multilabel classification of drug-like molecules via max-margin conditional random fields. In P. Myllymäki, T. Roos, & T. Jaakkola (Eds.), Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010): 13-15 September, 2010, Helsinki, Finland (pp. 265-272). (HIIT Publications; Vol. 2010, No. 2). Helsinki Institute for Information Technology HIIT.

Su, Hongyu ; Heinonen, Markus ; Rousu, Juho. / Multilabel classification of drug-like molecules via max-margin conditional random fields. Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010): 13-15 September, 2010, Helsinki, Finland. editor / Petri Myllymäki ; Teemu Roos ; Tommi Jaakkola. Helsinki Institute for Information Technology HIIT, 2010. pp. 265-272 (HIIT Publications; 2).

@inproceedings{4de526041ecb4a00b6f8fde356ada150,

title = "Multilabel classification of drug-like molecules via max-margin conditional random fields",

abstract = "We present a multilabel learning approach for molecular classification, an important task in drug discovery. We use a conditional random field to model the dependencies between drug targets and discriminative training to separate correct multilabels from incorrect ones with a large margin. Efficient training of the model is ensured by conditional gradient optimization on the marginal dual polytope, using loopy belief propagation to find the steepest feasible ascent directions. In our experiments, the MMCRF method outperformed the support vector machine with state-of-the-art graph kernels on a dataset comprising of cancer inhibition potential of drug-like molecules against a large number cancer cell lines.",

author = "Hongyu Su and Markus Heinonen and Juho Rousu",

year = "2010",

language = "English",

series = "HIIT Publications",

publisher = "Helsinki Institute for Information Technology HIIT",

number = "2",

pages = "265--272",

editor = "Petri Myllym{\"a}ki and Teemu Roos and Tommi Jaakkola",

booktitle = "Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010)",

address = "Finland",

note = "European Workshop on Probabilistic Graphical Models, PGM ; Conference date: 13-09-2010 Through 15-09-2010",

}

Su, H, Heinonen, M & Rousu, J 2010, Multilabel classification of drug-like molecules via max-margin conditional random fields. in P Myllymäki, T Roos & T Jaakkola (eds), Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010): 13-15 September, 2010, Helsinki, Finland. HIIT Publications, no. 2, vol. 2010, Helsinki Institute for Information Technology HIIT, pp. 265-272, European Workshop on Probabilistic Graphical Models, Helsinki, Finland, 13/09/2010.

Multilabel classification of drug-like molecules via max-margin conditional random fields. / Su, Hongyu; Heinonen, Markus ; Rousu, Juho.
Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010): 13-15 September, 2010, Helsinki, Finland. ed. / Petri Myllymäki; Teemu Roos; Tommi Jaakkola. Helsinki Institute for Information Technology HIIT, 2010. p. 265-272 (HIIT Publications; Vol. 2010, No. 2).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Multilabel classification of drug-like molecules via max-margin conditional random fields

AU - Su, Hongyu

AU - Heinonen, Markus

AU - Rousu, Juho

N1 - Conference code: 5

PY - 2010

Y1 - 2010

N2 - We present a multilabel learning approach for molecular classification, an important task in drug discovery. We use a conditional random field to model the dependencies between drug targets and discriminative training to separate correct multilabels from incorrect ones with a large margin. Efficient training of the model is ensured by conditional gradient optimization on the marginal dual polytope, using loopy belief propagation to find the steepest feasible ascent directions. In our experiments, the MMCRF method outperformed the support vector machine with state-of-the-art graph kernels on a dataset comprising of cancer inhibition potential of drug-like molecules against a large number cancer cell lines.

AB - We present a multilabel learning approach for molecular classification, an important task in drug discovery. We use a conditional random field to model the dependencies between drug targets and discriminative training to separate correct multilabels from incorrect ones with a large margin. Efficient training of the model is ensured by conditional gradient optimization on the marginal dual polytope, using loopy belief propagation to find the steepest feasible ascent directions. In our experiments, the MMCRF method outperformed the support vector machine with state-of-the-art graph kernels on a dataset comprising of cancer inhibition potential of drug-like molecules against a large number cancer cell lines.

UR - https://helda.helsinki.fi/handle/10138/18633

M3 - Conference article in proceedings

T3 - HIIT Publications

SP - 265

EP - 272

BT - Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010)

A2 - Myllymäki, Petri

A2 - Roos, Teemu

A2 - Jaakkola, Tommi

PB - Helsinki Institute for Information Technology HIIT

T2 - European Workshop on Probabilistic Graphical Models

Y2 - 13 September 2010 through 15 September 2010

ER -

Su H, Heinonen M , Rousu J. Multilabel classification of drug-like molecules via max-margin conditional random fields. In Myllymäki P, Roos T, Jaakkola T, editors, Proceedings of The Fifth European Workshop on Probabilistic Graphical Models (PGM-2010): 13-15 September, 2010, Helsinki, Finland. Helsinki Institute for Information Technology HIIT. 2010. p. 265-272. (HIIT Publications; 2).

Multilabel classification of drug-like molecules via max-margin conditional random fields

Abstract

Publication series

Workshop

UN SDGs

Other files and links

Fingerprint

Cite this