Advanced search
1 file | 1.53 MB

Robust peptide-based models in quantitative proteomics

Ludger Goeminne (UGent) , Klaas Vandepoele (UGent) , Kris Gevaert (UGent) and Lieven Clement (UGent)
(2015)
Author
Organization
Abstract
Peptide level models for assessing differential proteomics outperform summarization-based methods in terms of sensitivity, specificity, accuracy and precision (Goeminne et al., 2015, submitted). However, the ordinary least squares (OLS) parameter estimator is prone to overfitting and suffers from missing peptides and outliers that are omnipresent in proteomics data. We propose a robust ridge estimator and adopt empirical Bayes to stabilize the variance. With the CPTAC spike-in study, we demonstrate that our robust peptide-based estimator further improves the sensitivity and specificity.
Keywords
statistics, biostatistics, data analysis, differential protein abundance, label-free quantification, differential proteomics, peptide-based linear model, robust ridge regression, M estimation, Huber weights, empirical Bayes variance estimation, tandem mass spectrometry, overfitting, outliers, missing values

Downloads

  • Goeminne Proteomic Forum2015 poster.pdf
    • full text
    • |
    • open access
    • |
    • PDF
    • |
    • 1.53 MB

Citation

Please use this url to cite or link to this publication:

Chicago
Goeminne, Ludger, Klaas Vandepoele, Kris Gevaert, and Lieven Clement. 2015. “Robust Peptide-based Models in Quantitative Proteomics.” In .
APA
Goeminne, L., Vandepoele, K., Gevaert, K., & Clement, L. (2015). Robust peptide-based models in quantitative proteomics. Presented at the Proteomic Forum 2015.
Vancouver
1.
Goeminne L, Vandepoele K, Gevaert K, Clement L. Robust peptide-based models in quantitative proteomics. 2015.
MLA
Goeminne, Ludger et al. “Robust Peptide-based Models in Quantitative Proteomics.” 2015. Print.
@inproceedings{8616167,
  abstract     = {Peptide level models for assessing differential proteomics outperform summarization-based methods in terms of sensitivity, specificity, accuracy and precision (Goeminne et al., 2015, submitted). However, the ordinary least squares (OLS) parameter estimator is prone to overfitting and suffers from missing peptides and outliers that are omnipresent in proteomics data. We propose a robust ridge estimator and adopt empirical Bayes to stabilize the variance. With the CPTAC spike-in study, we demonstrate that our robust peptide-based estimator further improves the sensitivity and specificity.},
  author       = {Goeminne, Ludger and Vandepoele, Klaas and Gevaert, Kris and Clement, Lieven},
  language     = {eng},
  location     = {Berlin},
  title        = {Robust peptide-based models in quantitative proteomics},
  year         = {2015},
}