Skip to main content

Table 2 Area under ROC curves from different methods for Exon array data

From: puma 3.0: improved uncertainty propagation methods for gene and transcript expression analysis

Methods

2 replicates

5 replicates

  

1

2

3

4

5

Average

 

t-test

RMA

0.8945

0.8909

0.9107

0.9346

0.9316

0.9118

0.9475

 

PLIER

0.8806

0.8852

0.9004

0.9084

0.9083

0.8937

0.9291

 

GME

0.9082

0.9044

0.9415

0.9544

0.9427

0.9287

0.9580

PPLR_1000

RMA

0.9243

0.9234

0.9385

0.9417

0.9387

0.9323

0.9489

 

GME

0.9208

0.9093

0.9365

0.9297

0.8969

0.9188

0.9447

*PPLR_10000

RMA

0.9227

0.9226

0.9419

0.9453

0.9432

0.9348

0.9492

 

GME

0.9353

0.9317

0.9474

0.9374

0.9324

0.9274

0.9503

IPPLR

RMA

0.9246

0.9301

0.9464

0.9468

0.9463

0.9382

0.9493

 

GME

0.9379

0.9391

0.9457

0.9597

0.9549

0.9475

0.9589

  1. Gene expression estimation methods are combined with different finding-DE-gene methods. PPLR and IPPLR require a level of uncertainty associated with expression estimation, and they are therefore combined with GME and RMA since these two methods can provide variance of gene expression measurements. For t-test we use only the point estimates of gene expression. PLIER provides only a point estimate for gene expression and we only evaluate it combining with t-test. The number after PPLR indicates the sample number used in the importance sampling of the algorithm. The best result for each comparison is highlighted in bold.