Skip to main content

Table 1 Estimated power from 2000 simulations from a multivariate normal distribution with 2, 3, or 4 partitions with 15, 30, and 45 omics variables

From: TreeKernel: interpretable kernel machine tests for interactions between -omics and clinical predictors with applications to metabolomics and COPD phenotypes

 

Test

TreeKernel

Univariate Simes

Principal Component F-test

15 omics variables

4-partition

 2 Active groups

Group 1

Group 2

Group 1

Group 2

Group 1

Group 2

 

0.563

0.214

0.673

0.334

0.154

0.107

 1 Active group

0.740

0.696

0.832

   

3-Partition

 2 Active groups

Group 1

Group 2

Group 1

Group 2

Group 1

Group 2

 

0.898

0.613

0.597

0.023

0.833

0.018

 1 Active group

0.978

0.843

0.918

   

2-Partition

 1 Active group

0.998

0.965

0.962

   

30 omics variables

      

4-Partition

 2 Active groups

Group 1

Group 2

Group 1

Group 2

Group 1

Group 2

 

0.633

0.338

0.298

0.099

0.443

0.326

 1 Active group

0.723

0.510

0.621

   

3-Partition

      

 2 Active groups

Group 1

Group 2

Group 1

Group 2

Group 1

Group 2

 

0.909

0.771

0.486

0.107

0.656

0.020

 1 Active group

0.980

0.703

0.766

   

2-Partition

 1 Active group

0.998

0.931

0.884

   

45 omics variables

      

4-Partition

 2 Active groups

Group 1

Group 2

Group 1

Group 2

Group 1

Group 2

 

0.706

0.691

0.219

0.109

0.359

0318

 1 Active group

0.704

0.358

0.508

   

3-Partition

      

 2 Active groups

Group 1

Group 2

Group 1

Group 2

Group 1

Group 2

 

0.912

0.881

0.347

0.017

0.551

0.016

 1 Active group

0.960

0.556

0.659

   

2-Partition

 1 active group

0.996

0.885

0.799

   
  1. Tests used five principal components for clustering. Bold cells indicate the top performance within the simulation. ‘Group 1’ and ‘Group 2’ refer to the two partitions where the outcome was related to the simulated pathway