Dependence of the maxent inference of 20 solvation energy parameters for the “4state̱reduced” decoy set on the native probability . (a) Correlation of the inferred parameters obtained at a given probability with those obtained at . (b) Correlation of the inferred parameters with the Cornette hydrophobic scale. (c) Rate of successful learning of the proteins in the decoy set.
The total number of proteins whose native state has the lowest energy (a) or is among the five lowest energy conformations (b) using the inferred solvation energy parameters from eight different learning sets as shown in column 2 of Table I, and on employing different scoring functions as described in the text. The horizontal solid line indicates the performance of the hydrophobic scale as shown in Table I. The points connected by broken lines correspond to the data in the seventh (a) and eighth (b) columns of Table II–V, as indicated, and are obtained with different scoring functions given by Eqs. (12)–(15), respectively.
Decoy sets used in the learning. The rate of successful prediction (fifth and sixth columns) presents the fraction of proteins whose native state has the lowest energy or is among five lowest energy conformations (rank ) when using the Cornette hydrophobic scale as the solvation energy parameters in Eq. (12).
Inference of the 20 solvation energy parameters using maxent with . The rate of successful learning (third column) shows the number of proteins whose native state is ranked first (having the lowest energy) among the decoys in the learning set over the total number of proteins in the training set. The correlations of the inferred solvation energy parameters with the Cornette hydrophobic scale are given in the fourth column. The inferred energy parameters are then used to predict the native state of proteins which are present in the combined decoy sets 1–6 but not present in the training set. The rate of successful prediction (fifth and sixth columns) is defined as fraction of proteins successfully predicted using the inferred parameters. The prediction is considered as successful in two cases: When the native state has the lowest energy and when the native state is found among five conformations with the lowest energy . The total rates of success are determined by applying the inferred energy parameters on all proteins in the “combined sets 1–6” and are given in the seventh and eighth columns.
Inference of the 40 solvation energy parameters and in Eq. (13) using the maxent procedure with . The fourth column shows the correlations of the inferred solvation energy parameters with the Cornette hydrophobic scale.
Inference and results of testing of 40 parameters: 20 Solvation energy parameters and 20 contact energy parameters as defined in Eq. (14) were deduced using maxent with .
Inference of 60 parameters: 20 Solvation energy parameters and 40 energy parameters related to the bond angle and the torsional angles as defined in Eq. (15) were deduced using maxent with .
Values of the inferred solvation energy parameters for the 20 amino acids by the maxent method using the combined decoy sets 1–6 as the training set. The parameters shown are (second column) (third column), (fourth column), and (fifth column) as defined in Eqs. (12)–(15), respectively. The parameters in each column are normalized in such a way that their rms value is the same as that of the Cornette hydrophobic scale (shown in the sixth column).
Article metrics loading...
Full text loading...