Skip to main content

Table 5 Average FPR and TPR by unadjusted analysis on un-normalized/normalized simulated DM 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

0.0522

0.0115

0.0523

0.0115

-

-

-

-

DESeq2

0.0954

0.0305

0.0951

0.0304

-

-

-

-

EdgeR

0.0588

0.0130

0.0580

0.0133

-

-

-

-

LDM

0.0494

0.0097

0.0493

0.0098

0.0494

0.0097

0.0499

0.0097

Limma

0.0493

0.0098

0.0493

0.0101

-

-

-

-

Linear regression

0.0475

0.0085

0.0475

0.0085

0.0475

0.0085

0.0496

0.0101

MetagenomeSeq

-

-

-

-

-

-

0.1354

0.0552

Monocle

0.9463

0.9296

0.9452

0.9283

0.0481

0.0087

0.0501

0.0102

QRank

0.0489

0.0098

0.0496

0.0096

0.0489

0.0097

0.0491

0.0096

ZINQ-MinP

0.0468

0.0092

0.0468

0.0091

0.0466

0.0090

0.0478

0.0092

ZINQ-Cauchy

0.0522

0.0108

0.0523

0.0106

0.0524

0.0107

0.0523

0.0108

 

TPR

Corncob

0.3009

0.1678

0.3000

0.1675

-

-

-

-

DESeq2

0.2210 ∗

0.1095 ∗

0.2207 ∗

0.1093 ∗

-

-

-

-

EdgeR

0.1603

0.0642 ∗

0.1610

0.0652 ∗

-

-

-

-

LDM

0.1554

0.0646

0.1553

0.0646

0.1554

0.0646

0.2775

0.1530

Limma

0.2923

0.1647

0.2918

0.1646

-

-

-

-

Linear regression

0.1529

0.0611

0.1528

0.0611

0.1528

0.0610

0.2888

0.1619

MetagenomeSeq

-

-

-

-

-

-

0.3165 ∗

0.1955 ∗

Monocle

0.9610 ∗

0.9485 ∗

0.9603 ∗

0.9476 ∗

0.1537

0.0616

0.2901

0.1630

QRank

0.2318

0.1156

0.2316

0.1152

0.2325

0.1162

0.2253

0.1088

ZINQ-MinP

0.2419

0.1244

0.2414

0.1236

0.2422

0.1242

0.2391

0.1194

ZINQ-Cauchy

0.2820

0.1511

0.2814

0.1506

0.2819

0.1514

0.2785

0.1449

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