The utility of comparative models and the local model quality for protein crystal structure determination by Molecular Replacement

Table 1 Global model quality measures in the definition of model usefulness for MR

Predictor name	ALL_20	BACKBONE_20	ALL_IDEAL	BACKBONE_IDEAL	ALL_CA_IDEAL	BACKBONE_CA_IDEAL	MetaMQAP-evaluated	MetaMQAPclust-evaluated	POLYALA_20	POLYALA	MEAN (All Model Types)
Part A: Area under the ROC curve ( AUC)
RMSD	0.842	0.846	0.845	0.838	0.841	0.831	0.804	0.841	0.864	0.856	0.841
GDT_TS	0.851	0.852	0.849	0.839	0.847	0.831	0.821	0.848	0.857	0.856	0.845
seqId_stru	0.859	0.866	0.786	0.784	0.802	0.792	0.859	0.81	0.879	0.872	0.831
seqId_seq	0.856	0.861	0.767	0.764	0.785	0.772	0.855	0.8	0.878	0.872	0.821
fraction_of_gaps	0.67	0.698	0.697	0.702	0.688	0.682	0.694	0.697	0.708	0.712	0.695
aligned_columns	0.653	0.662	0.535	0.54	0.561	0.566	0.69	0.586	0.674	0.67	0.614
global_alignment score	0.816	0.823	0.715	0.724	0.729	0.728	0.833	0.749	0.826	0.815	0.776
target length	0.672	0.674	0.545	0.556	0.579	0.577	0.701	0.61	0.693	0.687	0.629
above_1.5	0.78	0.784	0.717	0.718	0.715	0.718	0.787	0.732	0.811	0.799	0.756
abone_0.5	0.739	0.748	0.687	0.691	0.69	0.692	0.781	0.702	0.784	0.776	0.729
above_-0.5	0.686	0.695	0.621	0.623	0.633	0.645	0.756	0.635	0.694	0.697	0.669
above_-1.5	0.638	0.646	0.565	0.565	0.581	0.592	0.709	0.582	0.645	0.648	0.617
below_-1.5	0.591	0.594	0.593	0.594	0.58	0.585	0.591	0.59	0.595	0.587	0.590
AmIgoMR	0.868	0.88	0.799	0.802	0.809	0.812	0.869	0.846	0.875	0.864	0.842
Part B: Optimal threshold value
RMSD	1.61	1.61	1.71	1.70	1.70	1.70	1.68	1.70	1.60	1.60	1.660
GDT_TS	79.68	79.68	78.11	77.71	78.25	78.93	79.68	78.86	79.68	79.68	79.02
seqId_stru	0.31	0.31	0.24	0.24	0.25	0.25	0.32	0.25	0.31	0.31	0.278
seqId_seq	31.05	31.05	24.02	24.02	25.07	27.06	33.05	27.06	35.05	31.05	28.84
AmIgoMR	0.29	0.29	0.18	0.17	0.20	0.20	0.29	0.19	0.39	0.29	0.249

Part A presents the accuracy of predictions of whether a model is suitable for MR. The AUC values are presented for all types of search models studied here. In addition, for each predictor also the averaged AUC value is shown. Part B shows the optimal threshold values for the predictors. Such a point reports to the value of a predictor for which the success of MR drops significantly. Predictors tested: RMSD and GDT_TS - global model quality computed by LGA program. seqId_stru and seqId_seq – sequence identity computer for structural alignment (TM-align) and sequence alignment (HHalign), respectively. The following parameters describe target-template sequence alignment made by the HHalign program: fraction_of_gaps, aligned_columns, global_alignment_score, target length, above_1.5, above_0.5, above_-0.5, above_-1.5, below_-1.5. AmIgoMR – a predictor based on the logistic regression on abovementioned HHalign parameters.

ISSN: 1471-2105