Use of biplots and partial least squares regression in microarray data analysis for assessing association between genes involved in different biological pathways

  • Authors:

  • Niccoló Bassani;Federico Ambrogi;Danila Coradini;Elia Biganzoli

  • Affiliations:

  • University of Milano, MI;University of Milano, MI;University of Milano, MI;University of Milano and Unit of Medical Statistics, Biometry and Bioinformatics, National Cancer Institute, Milano, MI

  • Venue:
  • CIBB'10 Proceedings of the 7th international conference on Computational intelligence methods for bioinformatics and biostatistics
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

Microarrays are widely used to study expression profiles for thousand of transcripts simultaneously and to explore interrelationships between sets of genes. Visualization techniques and Partial Least Squares (PLS) regression have thus gained relevance in genomic. Biplots provide an aid to understand relationships between genes and samples and among genes, whereas passive projections of variables are helpful for understanding conditional relationships between sets of genes to be quantitatively evaluated via PLS regression. 62 genes involved in loss of cell polarity and 8 involved in Epithelial-Mesenchymal Transition (EMT), were selected from a study on 49 mesothelioma samples, and analysis considered EMT genes as conditioning and polarity genes as conditioned variables. PLS regression results are consistent with the PCA-based biplot of EMT genes and with passive projections of polarity genes. Future work will address sparsity in PCA and PLS regression. PLS path modeling will be considered after specification of a detailed dependency network.