Evaluating Microarray-based Classifiers: An Overview

Boulesteix, Anne‐Laure; Strobl, Carolin; Augustin, Thomas; Däumer, Martin

doi:10.4137/cin.s408

Cited by 93 publications

(87 citation statements)

References 113 publications

(161 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The characteristics of the two datasets are represented in table 1, the number of iterations used to compute the test error rate with a Monte Carlo Cross Validation algorithm [3] is fixed to B = 50. …”

Section: Resultsmentioning

confidence: 99%

Graph-constrained discriminant analysis of functional genomics data

Guillemot

Brusquet

Tenenhaus³

et al. 2008

2008 IEEE International Conference on Bioinformatics and Biomeidcine Workshops

View full text Add to dashboard Cite

Classification studies from microarray data have proved useful in tasks like predicting patient class. At the same time, more and more biological information about gene regulation networks has been gathered mainly in the form of graph. Incorporating the a priori biological information encoded by graphs turns out to be a very important issue to increase classification performance. We present a method to integrate information from a network topology into a classification algorithm: the graphConstrained Discriminant Analysis (gCDA). We applied our algorithm to simulated and real data and show that it performs better than a linear Support Vector Machines classifier.

show abstract

Section: Resultsmentioning

confidence: 99%

Graph-constrained discriminant analysis of functional genomics data

Guillemot

Brusquet

Tenenhaus³

et al. 2008

2008 IEEE International Conference on Bioinformatics and Biomeidcine Workshops

View full text Add to dashboard Cite

show abstract

“…However, the choice of filtering methods and their tuning are still among open questions in high dimensional classification with PLS. The literature on variable filtering is rich with many pros and cons for both univariate and multivariate filtering approaches (Boulesteix et al 2008). Boulesteix (2004) observed that PLSDA reached best classification accuracy with more than one PLS component and suggested that Proof.…”

Section: Discussionmentioning

confidence: 99%

Sparse Partial Least Squares Classification for High Dimensional Data

Chung¹,

Keleş²

2010

Statistical Applications in Genetics and Molecular Biology

161

View full text Add to dashboard Cite

Partial least squares (PLS) is a well known dimension reduction method which has been recently adapted for high dimensional classification problems in genome biology. We develop sparse versions of the recently proposed two PLS-based classification methods using sparse partial least squares (SPLS). These sparse versions aim to achieve variable selection and dimension reduction simultaneously. We consider both binary and multicategory classification. We provide analytical and simulation-based insights about the variable selection properties of these approaches and benchmark them on well known publicly available datasets that involve tumor classification with high dimensional gene expression data. We show that incorporation of SPLS into a generalized linear model (GLM) framework provides higher sensitivity in variable selection for multicategory classification with unbalanced sample sizes between classes. As the sample size increases, the two-stage approach provides comparable sensitivity with better specificity in variable selection. In binary classification and multicategory classification with balanced sample sizes, the two-stage approach provides comparable variable selection and prediction accuracy as the GLM version and is computationally more efficient.

show abstract

“…response to treatment or disease course, and molecular diagnosis [21] as well as an algorithm-oriented approach to extracting the variables responsible for class separation and prediction. Common univariate methods like t-tests and Wilcoxon tests "are fast and conceptually simple.…”

Section: Classification With Variable Selectionmentioning

confidence: 99%

“…Common univariate methods like t-tests and Wilcoxon tests "are fast and conceptually simple. However, they do not take correlations and interactions between variables into consideration, resulting in a subset of variables that may not be optimal for classification" [21] . Multivariate variable selection approaches, on the other hand, recognize that the subset of variables with best univariate discrimination power are not necessarily the best subset of classification variables, and try to determine which combinations of variables yield high prediction accuracies.…”

Section: Classification With Variable Selectionmentioning

confidence: 99%

A classifier driven approach to find biomarkers for affective disorders from transcription profiles in blood

Mazin

Tamm

Antonijevic³

et al. 2016

apm

View full text Add to dashboard Cite

Gene expression profiles in blood are increasingly being used to identify biomarkers for different affective disorders. We have selected a set of 29 genes to generate expression profiles for healthy control subjects as well as for patients diagnosed with acute post-traumatic stress disorder (PTSD) and with borderline personality disorder (BPD). Measurements were performed by quantitative polymerase chain reaction (qPCR). Using the actual data in an anonymous form we constructed a series of artificial data sets with known gene expression profiles. These sets were used to test 14 classification algorithms and feature selection methods for their ability to identify the correct expression patterns. Application of the three most effective algorithms to the actual expression data showed that control subjects can be distinguished from BPD patients based on differential expression levels of the gene transcripts Gi2, GR and MAPK14, targets that may have links to stress related diseases. Controls can also be distinguished from acute PTSD patients by differential expression levels of the transcripts for ERK2 and RGS2 that are known to be associated with mood disorders and social anxiety. We conclude that it is possible to identify informative transcription profiles in blood samples from individuals with affective disorders. Keywords: feature selection, mental disorders, gene expressions, gene panel

show abstract

Evaluating Microarray-based Classifiers: An Overview

Cited by 93 publications

References 113 publications

Graph-constrained discriminant analysis of functional genomics data

Graph-constrained discriminant analysis of functional genomics data

Sparse Partial Least Squares Classification for High Dimensional Data

A classifier driven approach to find biomarkers for affective disorders from transcription profiles in blood

Contact Info

Product

Resources

About