Skip to main content

Table 4 Average FPR and TPR by adjusted analysis on un-normalized/normalized simulated OTU tables

From: Powerful and robust non-parametric association testing for microbiome data via a zero-inflated quantile approach (ZINQ)

Sample size = 531

        
 

Count

Rarefaction

TSS

CSS

α-level

0.05

0.01

0.05

0.01

0.05

0.01

0.05

0.01

 

FPR

Corncob

0.1077

0.0498

0.0919

0.0400

-

-

-

-

DESeq2

0.0921

0.0395

0.0779

0.0312

-

-

-

-

EdgeR

0.1034

0.0415

0.0893

0.0331

-

-

-

-

LDM

0.0501

0.0096

0.0499

0.0096

0.0501

0.0096

0.0489

0.0095

Limma

0.0561

0.0128

0.0719

0.0203

-

-

-

-

Linear regression

0.0475

0.0085

0.0469

0.0083

0.0472

0.0083

0.0488

0.0098

MetagenomeSeq

-

-

-

-

-

-

0.1539

0.0759

Monocle

0.7261

0.6695

0.6493

0.5839

0.0486

0.0086

0.0501

0.0102

QRank

0.0493

0.0101

0.0499

0.0100

0.0503

0.0098

0.0496

0.0099

ZINQ-MinP

0.0483

0.0094

0.0484

0.0096

0.0488

0.0096

0.0472

0.0091

ZINQ-Cauchy

0.0533

0.0107

0.0535

0.0106

0.0539

0.0104

0.0530

0.0109

 

TPR

Corncob

0.4544 ∗

0.3093 ∗

0.4018 ∗

0.2615 ∗

-

-

-

-

DESeq2

0.3289 ∗

0.2346 ∗

0.2859 ∗

0.1912 ∗

-

-

-

-

EdgeR

0.4046 ∗

0.2782 ∗

0.3653 ∗

0.2395 ∗

-

-

-

-

LDM

0.3283

0.1850

0.3094

0.1677

0.3283

0.1850

0.4150

0.2700

Limma

0.3981

0.2636 ∗

0.3701 ∗

0.2369 ∗

-

-

-

-

Linear regression

0.3358

0.1867

0.3030

0.1573

0.3214

0.1735

0.4080

0.2735

MetagenomeSeq

-

-

-

-

-

-

0.4900 ∗

0.3731 ∗

Monocle

0.8637 ∗

0.8275 ∗

0.8055 ∗

0.7579 ∗

0.3251

0.1766

0.4107

0.2761

QRank

0.3887

0.2346

0.2981

0.1641

0.3634

0.2160

0.3593

0.2117

ZINQ-MinP

0.4437

0.2733

0.3452

0.1945

0.4188

0.2535

0.4176

0.2519

ZINQ-Cauchy

0.4919

0.3156

0.3941

0.2333

0.4666

0.2943

0.4627

0.2919

  1. ∗: power of a method that inflates type I error
  2. +: power of a method that deflates type I error
  3. Results by the various methods on un-normalized/normalized simulated OTU table generated from the proposed two-part quantile model fitted on CARDIA data, including the average FPR and average TPR over 1000 runs according to significance cutoffs 0.05 and 0.01