Table 3

Genome-wide performance comparison among the Segal method, N-score, NuPoP, the Random method, the HMM method, DLaNe based on twelve individual structural features and the meta DLaNe method combing six features with the cutoff L = 35

Structural features

Prediction performance


Se

Sp

F-measure

Improvement(%)


DNA denaturation

0.703

0.412

0.520

18.47


Propeller twist

0.699

0.409

0.516

17.68


DNA-bending stiffness

0.702

0.408

0.516

17.68


Duplex disrupt energy

0.625

0.394

0.483

10.21


Bendability

0.623

0.391

0.480

9.56


Z-DNA

0.702

0.411

0.518

18.23


Stacking energy

0.695

0.408

0.514

17.25


Duplex free energy

0.689

0.404

0.509

16.15


Aphilicity

0.675

0.403

0.505

15.09


B-DNA twist

0.654

0.384

0.484

10.34


Protein-DNA twist

0.652

0.381

0.481

9.67


Protein deformation

0.526

0.353

0.422

-3.66


Meta DLaNe method

0.734

0.457

0.563

28.45


HMM method

0.723

0.445

0.551

25.63


Segal method

0.474

0.408

0.439

0.00


NuPoP method

0.356

0.489

0.412

-6.04


N-score method

0.317

0.439

0.368

-16.05


Random method

0.346

0.346

0.346

-21.10


The accuracy of a predicted nucleosome position is measured by sensitivity (Se), specificity (Sp) and F-measure. The improvement of F-measure is computed by comparing with the performance of the pioneer method (Segal method). For DLaNe method, the performance of all 12 structural features is shown in two part, according to their relationship with nucleosome occupancy

Gan et al. BMC Bioinformatics 2012 13:49   doi:10.1186/1471-2105-13-49

Open Data