Skip to main content

Table 8 Simulation results.

From: Comparison of lists of genes based on functional profiles

Onto.

s

n

m

A and B gene lists

Testing procedure

Pr{rejectH0} (true H0)

Pr{rejectH0} (false H0)

Reference: [25]

  

MF

88

69

52

AB =

Class-by-class

0.0012

0.3903

     

Chi-square

0.0334

1

     

New global

0.0469

1

     

Additional signif. classes

0.04585

0.697

BP

1602

372

328

AB =

Class-by-class

0.002

1

     

Chi-square

0.162

1

     

New global

0.042

1

     

Additional signif. classes

0.042

0

CC

298

305

336

AB =

Class-by-class

0.0042

1

     

Chi-square

0.0775

1

     

New global

0.0389

1

     

Additional signif. classes

0.0374

0

References: [27] and [28]

  

MF

88

110

99

AB

Class-by-class

0.0028

0.0729

  

k = 46

  

Chi-square

0.0341

0.998

     

New global

0.0428

0.7281

     

Additional signif. classes

0.0409

0.659

BP

1722

858

651

AB

Class-by-class

0.003

0.351

  

k = 318

  

Chi-square

0.152

1

     

New global

0.056

0.997

     

Additional signif. classes

0.055

0.646

CC

394

897

679

AB

Class-by-class

0.0076

0.9982

  

k = 354

  

Chi-square

0.0883

1

     

New global

0.0625

0.9999

     

Additional signif. classes

0.0599

0.0018

  1. Probability of rejecting the null hypothesis of equality of profiles at a nominal 5% significance level in different scenarios associated with real case studies at level 10 in the GO. In the column "testing procedure", "Class-by-class" stands for declaring global differences (i.e. rejecting the null hypothesis of profile equality) if at least one significant class is detected in a class-by-class analysis with correction for testing multiplicity; "Chi-square" stands for the classical chi-square test of homogeneity; "New global" stands for the global test presented in this paper and, finally, "Additional signif. classes" stands for step 3 in the algorithm proposed in the methods section, i.e. proportion of simulation replicates where additional significant classes were detected when a class-by-class analysis was unable of any detection.