Supplementary Figure 9: Evaluation of structure quality statistics for 20 ensembles of EC-NMR structures.
From: Protein structure determination by combining sparse NMR data with evolutionary couplings
The 20 ensembles of EC NMR structures generated in the sensitivity analysis outlined Supplementary Table 6 were assessed using structural quality metrics of the Protein Structure Validation Software Suite (PSVS), including NMR DP scores. The top panel plots backbone RMSD to reference structures (black data points) and the Number of Reliable EC Pairs (red data points), as defined in the On Line Methods, for each of the 20 EC-NMR analyses carried out using different amounts of sequence data (Neff/L). The bottom panel plots DP (left axis) or knowledge-base Z score (right axis) computed with the PSVS software. The metrics include the NMR DP score (red circles), ProCheck analysis of backbone dihedral angle distributions (Procheck, green boxes) and distributions for all backbone and sidechain dihedral angels (Procheck_all, blue diamonds), Veify3D (magenta triangles), ProsaII (grey triangles), and Molprobity atomic clash scores (black asterix). Z scores are normalized against a collection of high-resolution X-ray crystal structures. “Reliable” structures (backbone RMSD < 3.5 Å from X-ray crystal structure), obtained using multiple-sequence alignments of Neff/L > 5, all have DP score > ~ 0.73 and knowledge-base Z scores > -2 (upper right quadrant), while less accurate structures all have DP < ~ 0.73 and Z-scores for Verify3D < -2. 19 ensembles were generated using different amounts of evolutionary sequence data, and the 20th ensemble was generated using no evolutionary sequence data.
