gms | German Medical Science

64. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

08. - 11.09.2019, Dortmund

Article

Send article

Zero-sum regression in action: A prognostic miRNA Signature in DLBCL

Meeting Abstract

Search Medline for

  • Gunther Glehr - University of Regensburg, Institute of Functional Genomics, Statistical Bioinformatics, Regensburg, Germany
  • Carmen Nordmo - II. Medical Department, University Hospital Würzburg, Würzburg, Germany
  • Hilka Rauert-Wunderlich - Institute of Pathology, University of Würzburg, Würzburg, Germany; Comprehensive Cancer Centre (CCC) Mainfranken, Würzburg, Germany
  • Michael Altenbuchinger - Universität Regensburg, Regensburg, Germany
  • Andreas Rosenwald - Institute of Pathology, University of Würzburg, Würzburg, Germany; Comprehensive Cancer Centre (CCC) Mainfranken, Würzburg, Germany
  • Rainer Spang - Universität Regensburg, Institute of Functional Genomics, Regensburg, Germany

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 64. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Dortmund, 08.-11.09.2019. Düsseldorf: German Medical Science GMS Publishing House; 2019. DocAbstr. 267

doi: 10.3205/19gmds072, urn:nbn:de:0183-19gmds0721

Published: September 6, 2019

© 2019 Glehr et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License. See license information at http://creativecommons.org/licenses/by/4.0/.

Outline

Text

OMICs data sets need preprocessing before analysis. This intrinsically includes the use of reference points like the mean expression of all features, a defined spike-in or the value of housekeeper genes.

Reference points have two major drawbacks: Measuring platforms become incomparable [1] and noise of the reference point is added to a predictive model [2].

The concept of zero-sum signatures, introduced by [3] with extensions from [1] and [2], enables that a predicted response is free of reference points.

If all possible, unique log-ratios of measurements are used in a LASSO-penalized regression, such signatures directly emerge. However, the feature space expands from p measurements to p choose 2 new features.

Here we use expression levels of 800 micro-RNAs (miRNAs), measured with the NanoString nCounter miRNA system. 228 DLBCL specimens were used to find a predictive signature on all high-count log-ratios for overall and progression free survival. We show that, besides the zero-sum property, the found log-ratio features are predictive but the corresponding single measurements are not.

The authors declare that an ethics committee vote is not required.

This contribution has already been published: [4]

Outline

References

1.
Altenbuchinger M, Rehberg T, Zacharias HU, Stämmler F, Dettmer K, Weber D, Hiergeist A, Gessner A, Holler E, Oefner PJ, Spang R. Reference point insensitive molecular data analysis. Bioinformatics. 2016 Sep 15;33(2):219-26. DOI: 10.1093/bioinformatics/btw598 External link
2.
Bates S, Tibshirani R. Log-ratio lasso. Scalable, sparse estimation for log-ratio models. Biometrics. 2018 Nov 2. DOI: 10.1111/biom.12995 External link
3.
Lin W, Shi P, Feng R, Li H. Variable selection in regression with compositional covariates. Biometrika. 2014;101(4):785–797. DOI: 10.1093/biomet/asu031 External link
4.
Glehr G. Zero-sum regression in action: A prognostic miRNA Signature in DLBCL. In: Workshop on Computational Models in Biology and Medicine; 2019 March 7-8; BRICS, Braunschweig. 2019. p. 50. Available from: http://www.biometrische-gesellschaft.de/fileadmin/AG_Daten/MethodenBioinformatik/PDFs/program_workshop_2019.pdf External link