Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Table 1.

Summary of four datasets used in this study.

More »

Table 1 Expand

Table 2.

Comparison of the potential predictive capability of different features.

More »

Table 2 Expand

Figure 1.

Distribution comparison of network-based features of heme-binding and non-binding residues.

(a) Degree; (b) Closeness; (c) Betweenness; (d) Clustering Coefficient. For each topological feature, the z-scores of binding and non-binding residues in Dataset 1 are divided into high, medium and low score sections, and then the fraction occupied by each section was calculated.

More »

Figure 1 Expand

Figure 2.

Comparison of network-based features of heme-binding and non-binding residues based on residue type.

(a) Degree; (b) Closeness; (c) Betweenness; (d) Clustering Coefficient. For each topological feature, we compared the means of the z-scores of binding and non-binding residues in Dataset 1 for different residue types.

More »

Figure 2 Expand

Figure 3.

The ROC curves of different predictors tested on Dataset 1.

Seven SVM-based predictors were built in terms of different feature combination (N/G/C = Network/Geometry/Conservation) and evaluated by a 5-fold cross-validation on Dataset 1.

More »

Figure 3 Expand

Table 3.

Performance of different predictors tested on Dataset 1.

More »

Table 3 Expand

Figure 4.

The ROC curves of baseline model and HemeNet tested on Dataset 3.

The baseline model was the predictor considering only conventional features, while HemeNet incorporated network-based features into the baseline model.

More »

Figure 4 Expand

Table 4.

Performance of baseline model and HemeNet tested on Dataset 3.

More »

Table 4 Expand

Table 5.

Comparison of network-based features of holo-apo heme protein pairs.

More »

Table 5 Expand

Figure 5.

Visualization of the surface of holo-apo protein pair (2ZDO:A-1XBW:D).

(a) Degree; (b) Closeness; (c) Betweenness; (d) Clustering Coefficient. The color of the surface was changed according to the z-scores of each network-based feature (red/white/yellow = high/medium/low). The heme molecule was shown as a cyan stick. The holo structures (2ZDO:A) were presented in the upper row, and the apo structures (1XBW:D) were presented in the lower row. This figure was produced by PyMOL (www.pymol.org).

More »

Figure 5 Expand

Figure 6.

The ROC curves of baseline model and HemeNet tested on Dataset 4.

(a) Holo structures; (b) Apo structures.

More »

Figure 6 Expand

Table 6.

Performance of baseline model and HemeNet tested on Dataset 4.

More »

Table 6 Expand

Figure 7.

The ROC curves of ConCavity and HemeNet tested on holo-apo protein pairs.

(a) Holo structures; (b) Apo structures.

More »

Figure 7 Expand

Table 7.

Performance comparison of ConCavity and HemeNet.

More »

Table 7 Expand