Table 1.
Carbonylation datasets involved in this paper.
Figure 1.
Change of average MCC values versus the number of mRMR features and different window sizes using 10-fold cross-validation (n = 6 to 12 only).
(A) K carbonylation site prediction, (B) R carbonylation site prediction, (C) T carbonylation site prediction and (D) P carbonylation site prediction.
Figure 2.
ROC curves of the method corresponding to K, R, T and P carbonylation site predictions using 10-fold cross-validation.
Figure 3.
The distribution of the four kinds of features in the K, R, T, and P optimal feature sets.
(A) The number of each kind of features in the optimal feature sets. (B) The average Maximum Relevance scores of the four kinds of features in the optimal feature sets.
Figure 4.
Two-Sample-logos of the position-specific composition of residues surrounding carbonylation and non-carbonylation sites.
It shows position-specific residues enriched and depleted in positive samples of (A) K carbonylation site prediction, (B) R carbonylation site prediction, (C) T carbonylation site prediction and (D) P carbonylation site prediction, respectively.
Figure 5.
The average hydrophobicity at each position (excluding the carbonylation site itself) around carbonylation and non-carbonylation sites.
(A) K carbonylation site prediction, (B) R carbonylation site prediction, (C) T carbonylation site prediction and (D) P carbonylation site prediction.