Statistical integration of p-values for enhancing discovery of radiotoxicity gene signatures

Anna Papiez, Sylwia Kabacik, Christophe Badie, Simon Bouffler, Joanna Polanska

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

The aim of this study is to verify the effectiveness of a statistical integrative approach for merging expression data sets on the level of p-values. The data consist of two independent sets of expression levels for breast cancer patient lymphocyte tissue. The samples represent two groups: sensitive and resistant to the impact of ionizing radiation. Three approaches for integrating information derived from the two experiments to select a radiosensitivity gene signature were investigated: restrictive, non-statistical and integrative. Signature validity was assessed by verifying data separability using the support vector machine procedure and a logistic regression model selected using the likelihood ratio test to account for regularization. We demonstrated the value of additional information retained in the statistical method of gene selection based on combined p-values, which as the only logistic regression model in its optimal setting attained 100 % separability (AUC 86.2 % restrictive and 85.6 % non-statistical). Moreover, for the best support vector machine classifier, our p-value combination approach outperformed the restrictive and Arraymining methods (AUC 96.7 % versus 87.9 % and 94.6 % respectively). Further functional validation by signaling pathway research proved conjointly the biological relevance of the supreme gene signature.

Original languageEnglish
Title of host publicationBioinformatics and Biomedical Engineering - 3rd International Conference, IWBBIO 2015, Proceedings
EditorsFrancisco Ortuño, Ignacio Rojas
PublisherSpringer Verlag
Pages503-513
Number of pages11
ISBN (Print)9783319164823
DOIs
Publication statusPublished - 2015
Event3rd International Work Conference on Bioinformatics and Biomedical Engineering, IWBBIO 2015 - Granada, Spain
Duration: 15 Apr 201517 Apr 2015

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9043
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd International Work Conference on Bioinformatics and Biomedical Engineering, IWBBIO 2015
Country/TerritorySpain
CityGranada
Period15/04/1517/04/15

Keywords

  • Data integration
  • Feature selection
  • High-throughput
  • P-value combination
  • Radiosensitivity

Fingerprint

Dive into the research topics of 'Statistical integration of p-values for enhancing discovery of radiotoxicity gene signatures'. Together they form a unique fingerprint.

Cite this