Skip to main content

Table 4 Cluster accuracy and stability on yeast galactose data

From: Clustering gene-expression data with repeated measurements

(a) Cluster accuracy*

Algorithm

Similarity measure/model

Average

SD-weighted

CV-weighted

FITSS

True mean

IMM

Centroid linkage

Distance

0.968

0.849

0.802

0.968

0.159

NA

MCLUST-HC

NA

0.968

NA

NA

0.968

0.806

NA

Complete linkage

Distance

0.957

0.968

0.957

0.643

0.695

NA

Complete linkage

Spherical

NA

NA

NA

NA

NA

0.968

Complete linkage

Elliptical

NA

NA

NA

NA

NA

0.968

Centroid linkage

Correlation

0.942

0.807

0.753

0.942

0.942

NA

k-means

Correlation

0.871

0.640

0.827

NA

0.897

NA

Average linkage

Spherical

NA

NA

NA

NA

NA

0.897

Average linkage

Elliptical

NA

NA

NA

NA

NA

0.897

Average linkage

Distance

0.858

0.858

0.847

0.869

0.159

NA

Average linkage

Correlation

0.866

0.817

0.841

0.865

0.857

NA

k-means

Distance

0.857

0.857

0.767

NA

0.159

NA

Complete linkage

Correlation

0.677

0.724

0.730

0.503

0.744

NA

(b) Cluster stability†

Algorithm

Similarity measure/model

Average

SD-weighted

CV-weighted

FITSS

IMM

 

Complete linkage

Elliptical

NA

NA

NA

NA

0.998

 

Complete linkage

Spherical

NA

NA

NA

NA

0.991

 

Average linkage

Distance

0.820

0.985

0.914

0.650

NA

 

MCLUST-HC

NA

0.963

NA

NA

0.916

NA

 

Complete linkage

Distance

0.927

0.937

0.830

0.441

NA

 

Centroid linkage

Distance

0.893

0.924

0.841

0.893

NA

 

Average linkage

Spherical

NA

NA

NA

NA

0.923

 

k-means

Distance

0.905

0.867

0.798

NA

NA

 

Average linkage

Elliptical

NA

NA

NA

NA

0.895

 

Centroid linkage

Correlation

0.889

0.758

0.644

0.889

NA

 

Average linkage

Correlation

0.842

0.842

0.855

0.828

NA

 

k-means

Correlation

0.799

0.709

0.781

NA

NA

 

Complete linkage

Correlation

0.655

0.700

0.666

0.577

NA

 
  1. *Each entry shows the adjusted Rand index of the corresponding clustering approach with the four functional categories. A high adjusted Rand index represents close agreement with the external knowledge. †Each entry shows the average adjusted Rand index of the original clustering result with clusters from ten synthetic re-measured datasets. A high average adjusted Rand index means that clusters from synthetic remeasured data are in close agreement with clusters from the original dataset. For both parts of the table, the maximum adjusted Rand index of each row is shown in bold. The algorithms (rows) are sorted in descending order of the maximum average adjusted Rand index in each row. The external knowledge is not used in evaluating cluster stability.