Accurate prediction of DnaK-peptide binding via homology modelling and experimental data

Abstract

Molecular chaperones are essential elements of the protein quality control machinery that governs translocation and folding of nascent polypeptides, refolding and degradation of misfolded proteins, and activation of a wide range of client proteins. The prokaryotic heat-shock protein DnaK is the E. coli representative of the ubiquitous Hsp70 family, which specializes in the binding of exposed hydrophobic regions in unfolded polypeptides. Accurate prediction of DnaK binding sites in E. coli proteins is an essential prerequisite to understand the precise function of this chaperone and the properties of its substrate proteins. In order to map DnaK binding sites in protein sequences, we have developed an algorithm that combines sequence information from peptide binding experiments and structural parameters from homology modelling. We show that this combination significantly outperforms either single approach. The final predictor had a Matthews correlation coefficient (MCC) of 0.819 when assessed over the 144 tested peptide sequences to detect true positives and true negatives. To test the robustness of the learning set, we have conducted a simulated cross-validation, where we omit sequences from the learning sets and calculate the rate of repredicting them. This resulted in a surprisingly good MCC of 0.703. The algorithm was also able to perform equally well on a blind test set of binders and non-binders, of which there was no prior knowledge in the learning sets. The algorithm is freely available at http://limbo.vib.be.

Figure 1. DnaK binding to immobilised peptides attached to a cellulose-based membrane.

A) Group 2 peptides Membrane A, B) Group 2 peptides Membrane A bound to antibody only, indicating potential false positives (see supplementary Table 1 for all groups of peptide sequences and raw binding data).

Figure 2. DnaK bound to a substrate peptide.

Left: Substrate binding C-terminal domain of DnaK shown in cartoon style with bound heptapeptide NRLLLTG in stick style. Right: Detailed view of substrate binding in the beta sheet sandwich of DnaK. Molecular graphics created with YASARA (http://www.yasara.org) and Povray (http://www.povray.org).

Figure 3. ROC curves as calculated from the PSSMs of different DnaK-substrate structures.

The curves represent structures 1DKX (closed circle), 1DKY A (open square), 1DKY B (closed triangle) and 3 NMR structures from the same ensemble (open triangle, open circle, closed square).

Figure 4. ROC curves representing the performance of the different predictors before learning set optimization.

The graph visualizes the not cross-validated sequence-based predictor (open square), the not cross-validated sequence and structure-based predictor (closed square), the cross-validated sequence-based predictor (open circle), the cross-validated sequence and structure-based predictor (closed circle), the sequence based predictor on the independent validation set (open triangle) and the sequence and structure-based predictor on the independent validation set (closed triangle).

Figure 5. ROC curves representing the performance of the different predictors generated after running the learning set training algorithm.

The graph visualizes the not cross-validated sequence-based predictor (open square), the not cross-validated sequence and structure-based predictor (closed square), the cross-validated sequence-based predictor (open circle), the cross-validated sequence and structure-based predictor (closed circle), the sequence based predictor on the independent validation set (open triangle) and the sequence and structure-based predictor on the independent validation set (closed triangle).

Figure 6. Graphical representation of the final DnaK heptameric binding profile over all aminoacids and all residue positions.

The axes crossing point at score zero was chosen for convenience. Position 1 and 7 respectively represent the N-and and C-terminal end of the heptameric motif.

Figure 7. ROC performance comparison with a previously published algorithm.

A: Comparison of our DnaK predictor (closed square) with the previously published predicting algorithm from Rüdiger et al (open square) for the extended benchmark peptide set. The data point above 90% specificity (90.7%) with the best MCC in both predictors is shown as a dashed line. B: ROC curves comparison between our predictor (closed square) and the algorithm of Rüdiger et al (open square) for the validation set of peptides.