Skip to main content

Table 3 Comparison results of BGMM, BMM and GMM in applications to real data

From: A joint finite mixture model for clustering genes from independent Gaussian and beta distributed data

   

All

F

C

P

Dataset

Model

Criterion

Mean

Median

Mean

Median

Mean

Median

Mean

Median

 

BMM

4

0.2094

0.2246

0.3410

0.3453

0.3512

0.3552

0.3417

0.3404

Setrand 1

GMM

4

0.1958

0.1658

0.26719

0.3091

0.2925

0.3408

0.2747

0.3398

 

BGMM

4

0.1568

0.1047

0.2347

0.1826

0.2663

0.2536

0.2451

0.2287

  

BIC/AIC3

0.2071

0.1863

0.3261

0.3490

0.3408

0.3505

0.3331

0.3585

 

BMM

ICL

0.2013

0.2013

0.3080

0.3080

0.3631

0.3631

0.3594

0.3594

Setrand 2

 

AIC

0.1634

0.1505

0.2699

0.2499

0.2890

0.2772

0.2727

0.2427

 

GMM

4

0.1958

0.1658

0.2672

0.3091

0.2925

0.3408

0.2747

0.3398

 

BGMM

4

0.1436

0.0954

0.2198

0.2199

0.2453

0.2526

0.2311

0.2409

 

BMM

4

0.2204

0.2204

0.3748

0.3748

0.3799

0.3799

0.3769

0.3769

Setrand 3

GMM

4

0.1958

0.1658

0.2672

0.3091

0.2925

0.3408

0.2747

0.3398

 

BGMM

4

0.1466

0.1155

0.2623

0.2551

0.2838

0.3036

0.2714

0.2811

 

BMM

4

0.2407

0.2414

0.3228

0.3055

0.3575

0.3695

0.3300

0.3442

Set real

GMM

BIC/ICL

0.1973

0.1957

0.2799

0.2999

0.3040

0.3486

0.2883

0.3214

  

AIC/AIC3

0.1882

0.1708

0.2747

0.2917

0.3010

0.3325

0.2813

0.3103

 

BGMM

4

0.0987

0.0610

0.2455

0.2170

0.2894

0.2999

0.2558

0.2658

  1. Statistics shown in this table are the group average of the p-values of gene enrichment score (each p-value is the average of the p-values of all the ontology groups corresponding to the best clustering result selected by its corresponding model selection criterion). TFs of 'Setrand 1' to 'Setrand 3'were randomly chosen according to certain thresholds (details shown in the text), while TFs of 'Set real ' were carefully selected by BMM.
  2. Note: 'F', 'C' and 'P' stand for the three aspects of gene ontology; 'All' means all aspects are included. '4' represents that all the four criteria indicate the same clustering result. Statistics shown here are all rounded to 4 decimals.