Efficient Prediction of Nucleic Acid Binding Function from Low-resolution Protein Structures

Abstract

Structural genomics projects as well as ab initio protein structure prediction methods provide structures of proteins with no sequence or fold similarity to proteins with known functions. These are often low-resolution structures that may only include the positions of C a atoms. We present a fast and efficient method to predict DNA-binding proteins from just the amino acid sequences and low-resolution, C a -only protein models. The method uses the relative proportions of certain amino acids in the protein sequence, the asymmetry of the spatial distribution of certain other amino acids as well as the dipole moment of the molecule. These quantities are used in a linear formula, with coefficients derived from logistic regression performed on a training set, and DNA-binding is predicted based on whether the result is above a certain threshold. We show that the method is insensitive to errors in the atomic coordinates and provides correct predictions even on inaccurate protein models. We demonstrate that the method is capable of predicting proteins with novel binding site motifs and structures solved in an unbound state. The accuracy of our method is close to another, published method that uses all-atom structures, time-consuming calculations and information on conserved residues.

Frustration in protein–DNA binding influences conformational switching and target search kinetics

“…This trait allows the protein to alternate between one-dimensional sliding to hopping and dissociations from the DNA. Positively charged patches are common in DBPs and have unique features compared with positively charged patches found in protein-protein interfaces (11,12). For example, positively charged patches of DBPs are larger, have a higher α-helical content, greater hydrogen-bonding potency, and a higher degree of evolutionary conservation of positively charged residues than similar patches on non-DNA binding proteins (11).…”

“…This notion is supported by a greater dependence of the nonspecific interactions on salt concentration in comparison with specific protein-DNA complexes (8)(9)(10). Generally, DBPs have substantial regions of positive electrostatic patches at their DNA-binding interface that complement the negatively charged DNA (11,12). Negatively charged amino acids have a lower propensity in protein-DNA interfaces (13), though they might be observed and contribute to specificity by properly orienting the protein relative to the DNA (14) or interact with the DNA through a cation such as Mg 2þ , as observed in EcoRV endonuclease (15).…”

Gaussian Logic for Predictive Classification

“…[1], [3], [18], [20]). Szilágyi and Skolnick [19] created a logistic regression classifier based on 10 features including electrostatic dipole moment, proportions of charged amino acids Arg, Lys and Asp, spatial asymmetries of Arg and five more features not related to charged amino-acids: proportion of Ala and Gly and spatial asymmetry of Gly, Asn and Ser.…”

“…Classification was performed by comparing, for a tested protein, the likelihood-ratio of the two models (DNA-binding and non-DNA-binding) with a threshold selected on the training data. We estimated the accuracy of this method using 10-fold crossvalidation (always learning parameters and structure of the models and selecting the threshold and window length l w using only the data from training folds) on a dataset containing 138 DNA-binding proteins (PD138 [19]) and 110 non-DNAbinding proteins (NB110 [1]). The estimated accuracies (Gaussian Logic) are shown in Table 1.…”

Protein–DNA Electrostatics

“…Proteins that bind to DNA are most often positively charged. More precisely, positively charged patches are observed in the region which faces the DNA when the specific complex is formed, an effect which can be accounted for by evaluating the propensity of positive residues to occurs more frequently in a DNA-binding interface [14,15,65,66,67,68,25].…”