Medical Bioinformatics and e-Bioscience

---

Back to menu

.

Proteomics analysis

People

We have been involved in the experimental design and analysis of several SELDI-TOF proteomics experiments. Typically, the different steps in our analysis approach are experimental design: sample size, randomization of samples across multiple conditions quality control (QC): variety of QC plots and quantitative measures to aid in detecting problematic samples pre-processing: smoothing, baseline correction, normalization, alignment, peak detection and quantification unsupervised analysis: for example, using clustering and principal component analysis differential expression: detection of differentially expressed proteins for a variety of experimental designs (multiple groups, time series, inclusion of covariates, regression), emperical Bayes, correction for multiple testing, permutation tests classification: using a variety of statistical and machine learning models (discriminant analysis, decision trees, logistic regression, nearest centroid, support vector machines, neural networks, ...) in a double cross-validation scheme complete separation of training data used for estimating the parameters of a model and test data for estimating the accuracy of the model multiple estimates of classification accuracy to be able to assess its variance cross-validation on the training data to determine optimal values for hyperparameters of a model and to select features

Perry Moerland AntoineVanKampen

Selected publications
BC van Munster, MJ van Breemen, PD Moerland, D Speijer, SE de Rooij, CJG Pfrommer, M Levi, MW Hollmann, JM Aerts, AH Zwinderman, JC Korevaar (2009). Proteomic profiling of plasma and serum of elderly patients with postoperative delirium. Journal of Neuropsychiatry and Clinical Neurosciences, 21:284-91. PubMed

W Wegdam*, PD Moerland*, MR Buist, E Ver Loren van Themaat, B Bleijlevens, HCJ Hoefsloot, CG de Koster, JMFG Aerts (2009). Classification-based comparison of pre-processing methods for interpretation of mass spectrometry generated clinical datasets. Proteome Science, 7:19. PubMed *Contributed equally