Skip to main content

Table 8 Performances of different prediction tools on the three filtered test sets

From: PaPI: pseudo amino acid composition to score human protein-coding variants

Data Set

Tool

AUC

Balanced Accuracy

Sens

Spec

PPV

NPV

F-m

MCC

# 1

PaPI

.9218

.8575

.8518

.8631

.8989

.803

.8747

.7084

Carol

.9120

.8492

.821

.8774

.9054

.7742

.8611

.689

Provean

.8938

.8264

.7894

.8634

.892

.7415

.8375

.643

SIFT

.883

.8142

.7633

.8651

.8899

.7189

.8218

.6185

PolyPhen2

.9144

.8425

.8503

.8348

.8803

.796

.865

.6806

FATHMM

.8301

.7517

.6267

.8766

.8789

.6217

.7317

.502

LRT

.8455

.8249

.8009

.8488

.8833

.749

.8401

.6409

MutAssessor

.8899

.812

.7578

.8662

.89

.7144

.8186

.6141

# 2

PaPI

.9246

.863

.8623

.8637

.8989

.8169

.8802

.7209

Carol

.9121

.8442

.811

.8774

.9029

.7675

.8545

.6794

Provean

.8984

.8354

.8074

.8634

.8926

.7613

.8479

.6623

SIFT

.8836

.8091

.7532

.8651

.887

.7137

.8146

.6094

PolyPhen2

.9183

.8491

.8635

.8348

.8802

.813

.8717

.6957

FATHMM

.8355

.7603

.6441

.8766

.8801

.6366

.7438

.5187

LRT

.8506

.8317

.8147

.8488

.8834

.7651

.8477

.656

MutAssessor

.8923

.8134

.7606

.8662

.8888

.7202

.8197

.6178

# 3

PaPI

.9332

.8721

.8751

.8692

.9046

.8308

.8896

.7398

Carol

.9239

.8551

.8187

.8915

.9145

.7763

.8639

.7004

Provean

.9159

.8444

.8156

.8731

.9011

.7697

.8562

.6797

SIFT

.8911

.8166

.759

.8743

.8953

.7191

.8215

.6238

PolyPhen2

.9303

.8542

.8729

.8355

.8826

.8226

.8777

.7068

FATHMM

.8436

.7643

.6410

.8876

.8899

.6356

.7452

.527

LRT

.8682

.8408

.8289

.8527

.8886

.7786

.8577

.6744

MutAssessor

.8988

.8273

.7772

.8774

.8998

.7354

.8341

.6449

  1. Comparison of PaPI, PolyPhen2, SIFT, Carol, PROVEAN, FATHMM, LRT and MutationAssessor on the three test sets filtered for unpredictable variants by the other prediction tools. Area under the curve (AUC), balanced accuracy (sensitivity/2 + specificity/2), sensitivity (Sens), specificity (Spec), Positive Predictive Value (PPV), Negative Predictive Value (NPV), F-measure (F-m) and Matthews correlation coefficient (MCC) are reported for each method. Highest values for each set are marked in bold.