Projection to latent structures as a strategy for peptides microarray data analysis

D. S. Anisimov, S. V. Podlesnykh, E. A. Kolosova, D. N. Shcherbakov, V. D. Petrova, Stephen Johnston, A. F. Lazarev, N. N. Oskorbin, Andrei Chapoval, M. A. Ryazanov

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

Currently various microarrays platforms containing nucleotides, proteins, peptides, glycans and other molecules are used in biomedical research. Number and density of immobilized molecules on microarrays are constantly increasing. Microarray data handling requires optimization of methods for their analysis. Peptide microarrays data analysis has certain characteristics that require non-conventional statistical methods. In this paper we present the results of antibody repertoire analysis in breast cancer patients sera utilizing microchips containing 330,000 peptides. We investigated methods for space dimension reduction such as projective methods and methods for selection of informative features. We have shown that method of projection to latent structures can detect an effective data dimension, reduce overfitting of the model and increase the quality of object recognition. Accuracy of the experimental results was assessed with the ROC-curve; the best quality was achieved with three latent structures without normalization and reduction of total numbers of peptides.

Original languageEnglish (US)
Pages (from-to)435-445
Number of pages11
JournalMathematical Biology and Bioinformatics
Volume12
Issue number2
DOIs
StatePublished - 2017

Keywords

  • Clustering
  • Latent variables
  • Method of projection to latent structures
  • Microarrays
  • Normalization
  • Peptides
  • ROC

ASJC Scopus subject areas

  • Biomedical Engineering
  • Applied Mathematics

Fingerprint Dive into the research topics of 'Projection to latent structures as a strategy for peptides microarray data analysis'. Together they form a unique fingerprint.

Cite this