Drug susceptibility prediction against a panel of drugs using kernelized Bayesian multitask learning

Mehmet Gonen, Adam A. Margoliny

Research output: Contribution to journalArticle

34 Citations (Scopus)

Abstract

Motivation: Human immunodeficiency virus (HIV) and cancer require personalized therapies owing to their inherent heterogeneous nature. For both diseases, large-scale pharmacogenomic screens of molecularly characterized samples have been generated with the hope of identifying genetic predictors of drug susceptibility. Thus, computational algorithms capable of inferring robust predictors of drug responses from genomic information are of great practical importance. Most of the existing computational studies that consider drug susceptibility prediction against a panel of drugs formulate a separate learning problem for each drug, which cannot make use of commonalities between subsets of drugs. Results: In this study, we propose to solve the problem of drug susceptibility prediction against a panel of drugs in a multitask learning framework by formulating a novel Bayesian algorithm that combines kernel-based non-linear dimensionality reduction and binary classification (or regression). The main novelty of our method is the joint Bayesian formulation of projecting data points into a shared subspace and learning predictive models for all drugs in this subspace, which helps us to eliminate off-target effects and drug-specific experimental noise. Another novelty of our method is the ability of handling missing phenotype values owing to experimental conditions and quality control reasons. We demonstrate the performance of our algorithm via crossvalidation experiments on two benchmark drug susceptibility datasets of HIV and cancer. Our method obtains statistically significantly better predictive performance on most of the drugs compared with baseline single-task algorithms that learn drug-specific models. These results show that predicting drug susceptibility against a panel of drugs simultaneously within a multitask learning framework improves overall predictive performance over single-task learning approaches..

Original languageEnglish (US)
JournalBioinformatics
Volume30
Issue number17
DOIs
StatePublished - Sep 1 2014
Externally publishedYes

Fingerprint

Multi-task Learning
Bayesian Learning
Susceptibility
Drugs
Learning
Prediction
Viruses
Pharmaceutical Preparations
Quality control
Virus
Predictors
Cancer
Subspace
HIV
Experiments
Binary Regression
Benchmarking
Aptitude
Binary Classification
Pharmacogenetics

ASJC Scopus subject areas

  • Biochemistry
  • Molecular Biology
  • Computational Theory and Mathematics
  • Computer Science Applications
  • Computational Mathematics
  • Statistics and Probability
  • Medicine(all)

Cite this

Drug susceptibility prediction against a panel of drugs using kernelized Bayesian multitask learning. / Gonen, Mehmet; Margoliny, Adam A.

In: Bioinformatics, Vol. 30, No. 17, 01.09.2014.

Research output: Contribution to journalArticle

@article{964fb9155d01437ab697132056b6c0db,
title = "Drug susceptibility prediction against a panel of drugs using kernelized Bayesian multitask learning",
abstract = "Motivation: Human immunodeficiency virus (HIV) and cancer require personalized therapies owing to their inherent heterogeneous nature. For both diseases, large-scale pharmacogenomic screens of molecularly characterized samples have been generated with the hope of identifying genetic predictors of drug susceptibility. Thus, computational algorithms capable of inferring robust predictors of drug responses from genomic information are of great practical importance. Most of the existing computational studies that consider drug susceptibility prediction against a panel of drugs formulate a separate learning problem for each drug, which cannot make use of commonalities between subsets of drugs. Results: In this study, we propose to solve the problem of drug susceptibility prediction against a panel of drugs in a multitask learning framework by formulating a novel Bayesian algorithm that combines kernel-based non-linear dimensionality reduction and binary classification (or regression). The main novelty of our method is the joint Bayesian formulation of projecting data points into a shared subspace and learning predictive models for all drugs in this subspace, which helps us to eliminate off-target effects and drug-specific experimental noise. Another novelty of our method is the ability of handling missing phenotype values owing to experimental conditions and quality control reasons. We demonstrate the performance of our algorithm via crossvalidation experiments on two benchmark drug susceptibility datasets of HIV and cancer. Our method obtains statistically significantly better predictive performance on most of the drugs compared with baseline single-task algorithms that learn drug-specific models. These results show that predicting drug susceptibility against a panel of drugs simultaneously within a multitask learning framework improves overall predictive performance over single-task learning approaches..",
author = "Mehmet Gonen and Margoliny, {Adam A.}",
year = "2014",
month = "9",
day = "1",
doi = "10.1093/bioinformatics/btu464",
language = "English (US)",
volume = "30",
journal = "Bioinformatics",
issn = "1367-4803",
publisher = "Oxford University Press",
number = "17",

}

TY - JOUR

T1 - Drug susceptibility prediction against a panel of drugs using kernelized Bayesian multitask learning

AU - Gonen, Mehmet

AU - Margoliny, Adam A.

PY - 2014/9/1

Y1 - 2014/9/1

N2 - Motivation: Human immunodeficiency virus (HIV) and cancer require personalized therapies owing to their inherent heterogeneous nature. For both diseases, large-scale pharmacogenomic screens of molecularly characterized samples have been generated with the hope of identifying genetic predictors of drug susceptibility. Thus, computational algorithms capable of inferring robust predictors of drug responses from genomic information are of great practical importance. Most of the existing computational studies that consider drug susceptibility prediction against a panel of drugs formulate a separate learning problem for each drug, which cannot make use of commonalities between subsets of drugs. Results: In this study, we propose to solve the problem of drug susceptibility prediction against a panel of drugs in a multitask learning framework by formulating a novel Bayesian algorithm that combines kernel-based non-linear dimensionality reduction and binary classification (or regression). The main novelty of our method is the joint Bayesian formulation of projecting data points into a shared subspace and learning predictive models for all drugs in this subspace, which helps us to eliminate off-target effects and drug-specific experimental noise. Another novelty of our method is the ability of handling missing phenotype values owing to experimental conditions and quality control reasons. We demonstrate the performance of our algorithm via crossvalidation experiments on two benchmark drug susceptibility datasets of HIV and cancer. Our method obtains statistically significantly better predictive performance on most of the drugs compared with baseline single-task algorithms that learn drug-specific models. These results show that predicting drug susceptibility against a panel of drugs simultaneously within a multitask learning framework improves overall predictive performance over single-task learning approaches..

AB - Motivation: Human immunodeficiency virus (HIV) and cancer require personalized therapies owing to their inherent heterogeneous nature. For both diseases, large-scale pharmacogenomic screens of molecularly characterized samples have been generated with the hope of identifying genetic predictors of drug susceptibility. Thus, computational algorithms capable of inferring robust predictors of drug responses from genomic information are of great practical importance. Most of the existing computational studies that consider drug susceptibility prediction against a panel of drugs formulate a separate learning problem for each drug, which cannot make use of commonalities between subsets of drugs. Results: In this study, we propose to solve the problem of drug susceptibility prediction against a panel of drugs in a multitask learning framework by formulating a novel Bayesian algorithm that combines kernel-based non-linear dimensionality reduction and binary classification (or regression). The main novelty of our method is the joint Bayesian formulation of projecting data points into a shared subspace and learning predictive models for all drugs in this subspace, which helps us to eliminate off-target effects and drug-specific experimental noise. Another novelty of our method is the ability of handling missing phenotype values owing to experimental conditions and quality control reasons. We demonstrate the performance of our algorithm via crossvalidation experiments on two benchmark drug susceptibility datasets of HIV and cancer. Our method obtains statistically significantly better predictive performance on most of the drugs compared with baseline single-task algorithms that learn drug-specific models. These results show that predicting drug susceptibility against a panel of drugs simultaneously within a multitask learning framework improves overall predictive performance over single-task learning approaches..

UR - http://www.scopus.com/inward/record.url?scp=84907021110&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84907021110&partnerID=8YFLogxK

U2 - 10.1093/bioinformatics/btu464

DO - 10.1093/bioinformatics/btu464

M3 - Article

C2 - 25161247

AN - SCOPUS:84907021110

VL - 30

JO - Bioinformatics

JF - Bioinformatics

SN - 1367-4803

IS - 17

ER -