PNAbind: Structure-based prediction of protein-nucleic acid binding using graph neural networks.

The recognition and binding of nucleic acids (NAs) by proteins depends upon complementary chemical, electrostatic and geometric properties of the protein-NA binding interface. Structural models of protein-NA complexes provide insights into these properties but are scarce relative to models of unbound proteins. We present a deep learning approach for predicting protein-NA binding given the apo structure of a protein (PNAbind). Our method utilizes graph neural networks to encode spatial distributions of physicochemical and geometric properties of the protein molecular surface that are predictive of NA binding. Using global physicochemical encodings, our models predict the overall binding function of a protein and can discriminate between specificity for DNA or RNA binding. We show that such predictions made on protein structures modeled with AlphaFold2 can be used to gain mechanistic understanding of chemical and structural features that determine NA recognition. Using local encodings, our models predict the location of NA binding sites at the level of individual binding residues. Binding site predictions were validated against benchmark datasets, achieving AUROC scores in the range of 0.92-0.95. We applied our models to the HIV-1 restriction factor APOBEC3G and show that our predictions are consistent with experimental RNA binding data.

DeepPBS: Geometric deep learning for interpretable prediction of protein-DNA binding specificity.

Predicting specificity in protein-DNA interactions is a challenging yet essential task for understanding gene regulation. Here, we present Deep Predictor of Binding Specificity (DeepPBS), a geometric deep-learning model designed to predict binding specificity across protein families based on protein-DNA structures. The DeepPBS architecture allows investigation of different family-specific recognition patterns. DeepPBS can be applied to predicted structures, and can aid in the modeling of protein-DNA complexes. DeepPBS is interpretable and can be used to calculate protein heavy atom-level importance scores, demonstrated as a case-study on p53-DNA interface. When aggregated at the protein residue level, these scores conform well with alanine scanning mutagenesis experimental data. The inference time for DeepPBS is sufficiently fast for analyzing simulation trajectories, as demonstrated on a molecular-dynamics simulation of a Drosophila Hox-DNA tertiary complex with its cofactor. DeepPBS and its corresponding data resources offer a foundation for machine-aided protein-DNA interaction studies, guiding experimental choices and complex design, as well as advancing our understanding of molecular interactions.

Crystal Structures of Ternary Complexes of MEF2 and NKX2-5 Bound to DNA Reveal a Disease Related Protein-Protein Interaction Interface.

MEF2 and NKX2-5 transcription factors interact with each other in cardiogenesis and are necessary for normal heart formation. Despite evidence suggesting that these two transcription factors function synergistically and possibly through direct physical interactions, molecular mechanisms by which they interact are not clear. Here we determined the crystal structures of ternary complexes of MEF2 and NKX2-5 bound to myocardin enhancer DNA in two crystal forms. These crystal structures are the first example of human MADS-box/homeobox ternary complex structures involved in cardiogenesis. Our structures reveal two possible modes of interactions between MEF2 and NKX2-5: MEF2 and NKX bind to adjacent DNA sites to recognize DNA in cis; and MEF2 and NKX bind to different DNA strands to interact with each other in trans via a conserved protein-protein interface observed in both crystal forms. Disease-related mutations are mapped to the observed protein-protein interface. Our structural studies provide a starting point to understand and further study the molecular mechanisms of the interactions between MEF2 and NKX2.5 and their roles in cardiogenesis.

DNAproDB: an expanded database and web-based tool for structural analysis of DNA-protein complexes.

DNAproDB (https://dnaprodb.usc.edu) is a web-based database and structural analysis tool that offers a combination of data visualization, data processing and search functionality that improves the speed and ease with which researchers can analyze, access and visualize structural data of DNA-protein complexes. In this paper, we report significant improvements made to DNAproDB since its initial release. DNAproDB now supports any DNA secondary structure from typical B-form DNA to single-stranded DNA to G-quadruplexes. We have updated the structure of our data files to support complex DNA conformations, multiple DNA-protein complexes within a DNAproDB entry and model indexing for analysis of ensemble data. Support for chemically modified residues and nucleotides has been significantly improved along with the addition of new structural features, improved structural moiety assignment and use of more sequence-based annotations. We have redesigned our report pages and search forms to support these enhancements, and the DNAproDB website has been improved to be more responsive and user-friendly. DNAproDB is now integrated with the Nucleic Acid Database, and we have increased our coverage of available Protein Data Bank entries. Our database now contains 95% of all available DNA-protein complexes, making our tools for analysis of these structures accessible to a broad community.

Experimental maps of DNA structure at nucleotide resolution distinguish intrinsic from protein-induced DNA deformations.

Recognition of DNA by proteins depends on DNA sequence and structure. Often unanswered is whether the structure of naked DNA persists in a protein-DNA complex, or whether protein binding changes DNA shape. While X-ray structures of protein-DNA complexes are numerous, the structure of naked cognate DNA is seldom available experimentally. We present here an experimental and computational analysis pipeline that uses hydroxyl radical cleavage to map, at single-nucleotide resolution, DNA minor groove width, a recognition feature widely exploited by proteins. For 11 protein-DNA complexes, we compared experimental maps of naked DNA minor groove width with minor groove width measured from X-ray co-crystal structures. Seven sites had similar minor groove widths as naked DNA and when bound to protein. For four sites, part of the DNA in the complex had the same structure as naked DNA, and part changed structure upon protein binding. We compared the experimental map with minor groove patterns of DNA predicted by two computational approaches, DNAshape and ORChID2, and found good but not perfect concordance with both. This experimental approach will be useful in mapping structures of DNA sequences for which high-resolution structural data are unavailable. This approach allows probing of protein family-dependent readout mechanisms.

Expanding the repertoire of DNA shape features for genome-scale studies of transcription factor binding.

Uncovering the mechanisms that affect the binding specificity of transcription factors (TFs) is critical for understanding the principles of gene regulation. Although sequence-based models have been used successfully to predict TF binding specificities, we found that including DNA shape information in these models improved their accuracy and interpretability. Previously, we developed a method for modeling DNA binding specificities based on DNA shape features extracted from Monte Carlo (MC) simulations. Prediction accuracies of our models, however, have not yet been compared to accuracies of models incorporating DNA shape information extracted from X-ray crystallography (XRC) data or Molecular Dynamics (MD) simulations. Here, we integrated DNA shape information extracted from MC or MD simulations and XRC data into predictive models of TF binding and compared their performance. Models that incorporated structural information consistently showed improved performance over sequence-based models regardless of data source. Furthermore, we derived and validated nine additional DNA shape features beyond our original set of four features. The expanded repertoire of 13 distinct DNA shape features, including six intra-base pair and six inter-base pair parameters and minor groove width, is available in our R/Bioconductor package DNAshapeR and enables a comprehensive structural description of the double helix on a genome-wide scale.

Structure of the Forkhead Domain of FOXA2 Bound to a Complete DNA Consensus Site.

FOXA2, a member of the forkhead family of transcription factors, plays essential roles in liver development and bile acid homeostasis. In this study, we report a 2.8 Å co-crystal structure of the FOXA2 DNA-binding domain (FOXA2-DBD) bound to a DNA duplex containing a forkhead consensus binding site (GTAAACA). The FOXA2-DBD adopts the canonical winged-helix fold, with helix H3 and wing 1 regions mainly mediating the DNA recognition. Although the wing 2 region was not defined in the structure, isothermal titration calorimetry assays suggested that this region was required for optimal DNA binding. Structure comparison with the FOXA3-DBD bound to DNA revealed more major groove contacts and fewer minor groove contacts in the FOXA2 structure than in the FOXA3 structure. Structure comparison with the FOXO1-DBD bound to DNA showed that different forkhead proteins could induce different DNA conformations upon binding to identical DNA sequences. Our findings provide the structural basis for FOXA2 protein binding to a consensus forkhead site and elucidate how members of the forkhead protein family bind different DNA sites.

DNAproDB: an interactive tool for structural analysis of DNA-protein complexes.

Many biological processes are mediated by complex interactions between DNA and proteins. Transcription factors, various polymerases, nucleases and histones recognize and bind DNA with different levels of binding specificity. To understand the physical mechanisms that allow proteins to recognize DNA and achieve their biological functions, it is important to analyze structures of DNA-protein complexes in detail. DNAproDB is a web-based interactive tool designed to help researchers study these complexes. DNAproDB provides an automated structure-processing pipeline that extracts structural features from DNA-protein complexes. The extracted features are organized in structured data files, which are easily parsed with any programming language or viewed in a browser. We processed a large number of DNA-protein complexes retrieved from the Protein Data Bank and created the DNAproDB database to store this data. Users can search the database by combining features of the DNA, protein or DNA-protein interactions at the interface. Additionally, users can upload their own structures for processing privately and securely. DNAproDB provides several interactive and customizable tools for creating visualizations of the DNA-protein interface at different levels of abstraction that can be exported as high quality figures. All functionality is documented and freely accessible at http://dnaprodb.usc.edu.