Skip to main content

Table 2 Comparison of the properties of several geneset differential expression analysis methods

From: Functional Analysis: Evaluation of Response Intensities - Tailoring ANOVA for Lists of Expression Subsets

Author

Year

Hypothesis

Data used

Group statistic

Significance

Name

Properties

Individual statistic

Mootha et al

2003

Competitive

Individual statistic

ES (Running Sum)

Sample permutations

GSEA

Hybrid

Signal/noise Ratio

Subramanian et al

2005

Competitive

Individual statistic

ES (Running Sum)

Sample permutations

GSEA

Hybrid, asymmetrical

Individual correlation (r)

Keller et al

2007

Competitive

Individual statistic

ES (Running Sum)

Competitive theoretical model

Variant GSEA

Hybrid, symmetrical

*

Effron & Tibshirani

2007

Competitive

Individual statistic

ES (Running Sum)

Sample permutations

GSA

Restandardizaton

*

Pavlidis et al

2004

Competitive

Individual statistic

Log (p(g)) = mean[log[p(i)]]

Genes permutations

Ā 

Depends on the geneset size

Pearson correlation coef.

Tian et al

2005

Competitive

Individual statistic

Weighted mean

Genes permutations

Ā 

Standardization

Student t

Tian et al

2005

Self-contained

Individual statistic

Weighted mean

Sample permutations

Ā 

Standardization

Student t

Kim & Volsky

2005

Self-contained

Individual statistic

Mean

Normal distribution

PAGE

Central Limit Theorem

Fold-change

Effron & Tibshirani

2007

Self-contained

Individual statistic

Mean

Sample permutations

GSA

Unidirectional

+ Restandardization

Student t

Effron & Tibshirani

2007

Self-contained

Individual statistic

Maxmean

Sample permutations

GSA

Unidirectional (directional subset) + Restandardization

Student t

Effron & Tibshirani

2007

Self-contained

Individual statistic

Absmean

Sample permutations

GSA

Unidirectional (absolute value) + Restandardization

Student t

Dinu et al

2007

Self-contained

Expression data

Sum (d2)

Sample permutations

SAM-GS

Determination of S0

SAM d statistic

Goeman et al

2004

Self-contained

Expression data

Q(g) = mean(Q(i))

Permutation/Gamma/Asymptotic

GlobalTest

P (Y|X)

Q(i)

Mansmann & Meister

2005

Self-contained

Expression data

F

Sample permutations

GlobalAncova

P (X|Y)

Ā 

Berger et al

Unpublished

Self-contained

Expression data

F

Fisher F

ANOVA-2

Unidirectional

Ā 

Berger et al

Unpublished

Self-contained

Expression data

F*

Sample permutations/random data

FAERI

Bidirectional

Ā 
  1. The methods are grouped into categories. The upper and lower parts of the table list respectively the competitive and self-contained methods. The methods highlighted in bold rely on a two-step procedure. The methods in plain writing rely on a global analysis, which uses the expression data to compute the geneset statistic in one single step, based on multivariate models. Finally, ANOVA-2 and FAERI are shown in italics.