Table 5 Comparisons of the five-fold cross-validation performance values in the training set of two prediction systems based on three feature sets, combined by second-layer SVM.

From: The structure-based cancer-related single amino acid variation prediction

System

Feature set

Accuracy

Sensitivity

Specificity

MCC

Precision

F1 score

CanSavPrew

Sequence

0.7450

0.4350

0.8620

0.3202

0.5434

0.4832

Structure

0.6995

0.3946

0.8146

0.2176

0.4454

0.4185

Microenvironment

0.7184

0.3988

0.8391

0.2536

0.4832

0.4370

Combined

0.7983

0.4356

0.9357

0.4452

0.7199

0.5428

CanSavPrewm

Sequence

0.8471

0.6292

0.9293

0.5978

0.7706

0.6928

Structure

0.8022

0.4737

0.9262

0.4609

0.7078

0.5676

Microenvironment

0.8311

0.5774

0.9268

0.5509

0.7487

0.6520

Combined

0.8973

0.7837

0.9404

0.7382

0.8328

0.8075

  1. All predictions were optimized using MCC as the fitness function.
Back to article page