Table 5

Performance of the predictive models (M3 through M16), each with an individual or a combination of the newly added categories of features being excluded.

Features

SP

SE

ACC

CC


Histone Methylation Retained

All retained

0.9405

0.9257

0.9313

0.8302

Acetylation (M3)

0.9012

0.8965

0.9046

0.7852

Functional role (M4)

0.9302

0.9265

0.9210

0.8038

Nucleosome (M5)

0.9270

0.9250

0.9205

0.8024

Acetylation+Functional (M6)

0.8791

0.8903

0.8897

0.7632

Acetylation+Nucleosome (M7)

0.8698

0.8835

0.8826

0.7625

Functional+Nucleosome (M8)

0.9186

0.9116

0.9186

0.8012

All three (M9)

0.8685

0.8822

0.8786

0.7558


Histone Methylation Excluded

All but histone methylation

0.9318

0.5932

0.8575

0.6404

Acetylation (M10)

0.9670

0.2247

0.8001

0.3302

Functional (M11)

0.9092

0.5670

0.8312

0.6124

Nucleosome (M12)

0.9078

0.5660

0.8296

0.6076

Acetylation+Functional (M13)

0.9320

0.2279

0.7862

0.3236

Acetylation+Nucleosome (M14)

0.9266

0.2304

0.7641

0.3264

Functional+Nucleosome (M15)

0.8990

0.5519

0.8232

0.5924

All three (M16)

0.8972

0.2338

0.7352

0.3013


Specificity (SP), sensitivity (SE) and accuracy (ACC) are evaluated for binary classification, and correlation coefficient (CC) for regression models.

Zheng et al. BMC Medical Genomics 2013 6(Suppl 1):S13   doi:10.1186/1755-8794-6-S1-S13

Open Data