Automatic workflow for the classification of local DNA conformations

3dDNAscoreA: A scoring function for evaluation of DNA 3D structures

DNA molecules are vital macromolecules that play a fundamental role in many cellular processes and have broad applications in medicine. For example, DNA aptamers have been rapidly developed for diagnosis, biosensors, and clinical therapy. Recently, we proposed a computational method of predicting DNA 3D structures, called 3dDNA. However, it lacks a scoring function to evaluate the predicted DNA 3D structures, and so they are not ranked for users. Here, we report a scoring function, 3dDNAscoreA, for evaluation of DNA 3D structures based on a deep learning model ARES for RNA 3D structure evaluation but using a new strategy for training. 3dDNAscoreA is benchmarked on two test sets to show its ability to rank DNA 3D structures and select the native and near-native structures.

Accurate prediction of B-form/A-form DNA conformation propensity from primary sequence: A machine learning and free energy handshake

DNA carries the genetic code of life, with different conformations associated with different biological functions. Predicting the conformation of DNA from its primary sequence, although desirable, is a challenging problem owing to the polymorphic nature of DNA. We have deployed a host of machine learning algorithms, including the popular state-of-the-art LightGBM (a gradient boosting model), for building prediction models. We used the nested cross-validation strategy to address the issues of “overfitting” and selection bias. This simultaneously provides an unbiased estimate of the generalization performance of a machine learning algorithm and allows us to tune the hyperparameters optimally. Furthermore, we built a secondary model based on SHAP (SHapley Additive exPlanations) that offers crucial insight into model interpretability. Our detailed model-building strategy and robust statistical validation protocols tackle the formidable challenge of working on small datasets, which is often the case in biological and medical data.

•

A robust machine learning model to predict A- or B-DNA conformation

•

Outcome of machine learning model is explained with free energy values

•

Our approach works well under class imbalance and limited data constraints

The sequence in the genome of an organism encodes all the information of life. We combine a data-driven approach using machine learning (ML) and the results of free energy calculations to offer a fresh perspective on this long-standing problem of prediction of DNA conformation (A or B) from the sequence. We trained our ML model using sophisticated state-of-the art algorithms such as LightGBM along with a nested cross-validation strategy to overcome the common problems associated with data bias and overfitting when constrained by limited data size. Our study will serve the broader interest of researchers who are not only seeking accurate and reliable predictive models but also want to understand the physical and chemical origins behind the predictions.

Ion Binding Properties and Dynamics of the bcl-2 G-Quadruplex Using a Polarizable Force Field

G-quadruplexes (GQs) are topologically diverse, highly thermostable noncanonical nucleic acid structures that form in guanine-rich sequences in DNA and RNA. GQs are implicated in transcriptional and translational regulation and genome maintenance, and deleterious alterations to their structures contribute to diseases such as cancer. The expression of the B-cell lymphoma 2 (Bcl-2) antiapoptotic protein, for example, is under transcriptional control of a GQ in the promoter of the bcl-2 gene. Modulation of the bcl-2 GQ by small molecules is of interest for chemotherapeutic development but doing so requires knowledge of the factors driving GQ folding and stabilization. To develop a greater understanding of the electrostatic properties of the bcl-2 promoter GQ, we performed molecular dynamics simulations using the Drude-2017 polarizable force field and compared relevant outcomes to the nonpolarizable CHARMM36 force field. Our simulation outcomes highlight the importance of dipole-dipole interactions in the bcl-2 GQ, particularly during the recruitment of a bulk K⁺ ion to the solvent-exposed face of the tetrad stem. We also predict and characterize an "electronegative pocket" at the tetrad-long loop junction that induces local backbone conformational change and may induce local conformational changes at cellular concentrations of K⁺. These outcomes suggest that moieties within the bcl-2 GQ can be targeted by small molecules to modulate bcl-2 GQ stability.

Structural interpretation of the 31P NMR chemical shifts in thiophosphate and phosphate: key effects due to spin–orbit and explicit solvent

Structural interpretation of the ³¹P NMR shifts measured in different molecules including thiophosphate or phosphate group was obtained by means of theoretical calculations including the effects of geometry, molecular dynamics, solvent, relativistic effects and the effect of NMR reference.

A DNA structural alphabet provides new insight into DNA flexibility

DNA is a structurally plastic molecule, and its biological function is enabled by adaptation to its binding partners. To identify the DNA structural polymorphisms that are possible in such adaptations, the dinucleotide structures of 60 000 DNA steps from sequentially nonredundant crystal structures were classified and an automated protocol assigning 44 distinct structural (conformational) classes called NtC (for Nucleotide Conformers) was developed. To further facilitate understanding of the DNA structure, the NtC were assembled into the DNA structural alphabet CANA (Conformational Alphabet of Nucleic Acids) and the projection of CANA onto the graphical representation of the molecular structure was proposed. The NtC classification was used to define a validation score called confal, which quantifies the conformity between an analyzed structure and the geometries of NtC. NtC and CANA assignment were applied to analyze the structural properties of typical DNA structures such as Dickerson-Drew dodecamers, guanine quadruplexes and structural models based on fibre diffraction. NtC, CANA and confal assignment, which is accessible at the website https://dnatco.org, allows the quantitative assessment and validation of DNA structures and their subsequent analysis by means of pseudo-sequence alignment. An animated Interactive 3D Complement (I3DC) is available in Proteopedia at http://proteopedia.org/w/Journal:Acta_Cryst_D:2.

A DNA Structural Alphabet Distinguishes Structural Features of DNA Bound to Regulatory Proteins and in the Nucleosome Core Particle

We analyzed the structural behavior of DNA complexed with regulatory proteins and the nucleosome core particle (NCP). The three-dimensional structures of almost 25 thousand dinucleotide steps from more than 500 sequentially non-redundant crystal structures were classified by using DNA structural alphabet CANA (Conformational Alphabet of Nucleic Acids) and associations between ten CANA letters and sixteen dinucleotide sequences were investigated. The associations showed features discriminating between specific and non-specific binding of DNA to proteins. Important is the specific role of two DNA structural forms, A-DNA, and BII-DNA, represented by the CANA letters AAA and BB2: AAA structures are avoided in non-specific NCP complexes, where the wrapping of the DNA duplex is explained by the periodic occurrence of BB2 every 10.3 steps. In both regulatory and NCP complexes, the extent of bending of the DNA local helical axis does not influence proportional representation of the CANA alphabet letters, namely the relative incidences of AAA and BB2 remain constant in bent and straight duplexes.

Polarizable Force Field for DNA Based on the Classical Drude Oscillator: I. Refinement Using Quantum Mechanical Base Stacking and Conformational Energetics

Empirical force fields seek to relate the configuration of a set of atoms to its energy, thus yielding the forces governing its dynamics, using classical physics rather than more expensive quantum mechanical calculations that are computationally intractable for large systems. Most force fields used to simulate biomolecular systems use fixed atomic partial charges, neglecting the influence of electronic polarization, instead making use of a mean-field approximation that may not be transferable across environments. Recent hardware and software developments make polarizable simulations feasible, and to this end, polarizable force fields represent the next generation of molecular dynamics simulation technology. In this work, we describe the refinement of a polarizable force field for DNA based on the classical Drude oscillator model by targeting quantum mechanical interaction energies and conformational energy profiles of model compounds necessary to build a complete DNA force field. The parametrization strategy employed in the present work seeks to correct weak base stacking in A- and B-DNA and the unwinding of Z-DNA observed in the previous version of the force field, called Drude-2013. Refinement of base nonbonded terms and reparametrization of dihedral terms in the glycosidic linkage, deoxyribofuranose rings, and important backbone torsions resulted in improved agreement with quantum mechanical potential energy surfaces. Notably, we expand on previous efforts by explicitly including Z-DNA conformational energetics in the refinement.

Noncanonical α/γ Backbone Conformations in RNA and the Accuracy of Their Description by the AMBER Force Field

The sugar-phosphate backbone of RNA can exist in diverse rotameric substates, giving RNA molecules enormous conformational variability. The most frequent noncanonical backbone conformation in RNA is α/γ = t/t, which is derived from the canonical backbone by a crankshaft motion and largely preserves the standard geometry of the RNA duplex. A similar conformation also exists in DNA, where it has been extensively studied and shown to be involved in DNA-protein interactions. However, the function of the α/γ = t/t conformation in RNA is poorly understood. Here, we present molecular dynamics simulations of several prototypical RNA structures obtained from X-ray and NMR experiments, including canonical and mismatched RNA duplexes, UUCG and GAGA tetraloops, Loop E, the sarcin-ricin loop, a parallel guanine quadruplex, and a viral pseudoknot. The stability of various noncanonical α/γ backbone conformations was analyzed with two AMBER force fields, ff99bsc0χ_OL3 and ff99bsc0χ_OL3 with the recent εζ_OL1 and β_OL1 corrections for DNA. Although some α/γ substates were stable with seemingly well-described equilibria, many were unstable in our simulations. Notably, the most frequent noncanonical conformer α/γ = t/t was unstable in both tested force fields. Possible reasons for this instability are discussed. Our work reveals a potentially important artifact in RNA force fields and highlights a need for further force field refinement.

The benchmark of 31P NMR parameters in phosphate: a case study on structurally constrained and flexible phosphate

A benchmark for structural interpretation of the ³¹P NMR shift and the ²J_P,C spin–spin coupling in the phosphate group was obtained by means of theoretical calculations and measurements in diethylphosphate and 5,5-dimethyl-2-hydroxy-1,3,2-dioxaphosphinane 2-oxide.

Assessing the Current State of Amber Force Field Modifications for DNA

The utility of molecular dynamics (MD) simulations to model biomolecular structure, dynamics, and interactions has witnessed enormous advances in recent years due to the availability of optimized MD software and access to significant computational power, including GPU multicore computing engines and other specialized hardware. This has led researchers to routinely extend conformational sampling times to the microsecond level and beyond. The extended sampling time has allowed the community not only to converge conformational ensembles through complete sampling but also to discover deficiencies and overcome problems with the force fields. Accuracy of the force fields is a key component, along with sampling, toward being able to generate accurate and stable structures of biopolymers. The Amber force field for nucleic acids has been used extensively since the 1990s, and multiple artifacts have been discovered, corrected, and reassessed by different research groups. We present a direct comparison of two of the most recent and state-of-the-art Amber force field modifications, bsc1 and OL15, that focus on accurate modeling of double-stranded DNA. After extensive MD simulations with five test cases and two different water models, we conclude that both modifications are a remarkable improvement over the previous bsc0 force field. Both force field modifications show better agreement when compared to experimental structures. To ensure convergence, the Drew–Dickerson dodecamer (DDD) system was simulated using 100 independent MD simulations, each extended to at least 10 μs, and the independent MD simulations were concatenated into a single 1 ms long trajectory for each combination of force field and water model. This is significantly beyond the time scale needed to converge the conformational ensemble of the internal portions of a DNA helix absent internal base pair opening. Considering all of the simulations discussed in the current work, the MD simulations performed to assess and validate the current force fields and water models aggregate over 14 ms of simulation time. The results suggest that both the bsc1 and OL15 force fields render average structures that deviate significantly less than 1 Å from the average experimental structures. This can be compared to similar but less exhaustive simulations with the CHARMM 36 force field that aggregate to the ∼90 μs time scale and also perform well but do not produce structures as close to the DDD NMR average structures (with root-mean-square deviations of 1.3 Å) as the newer Amber force fields. On the basis of these analyses, any future research involving double-stranded DNA simulations using the Amber force fields should employ the bsc1 or OL15 modification.

DNATCO: assignment of DNA conformers at dnatco.org

The web service DNATCO (dnatco.org) classifies local conformations of DNA molecules beyond their traditional sorting to A, B and Z DNA forms. DNATCO provides an interface to robust algorithms assigning conformation classes called ntC to dinucleotides extracted from DNA-containing structures uploaded in PDB format version 3.1 or above. The assigned dinucleotide ntC classes are further grouped into DNA structural alphabet ntA, to the best of our knowledge the first DNA structural alphabet. The results are presented at two levels: in the form of user friendly visualization and analysis of the assignment, and in the form of a downloadable, more detailed table for further analysis offline. The website is free and open to all users and there is no login requirement.

Refinement of the Sugar-Phosphate Backbone Torsion Beta for AMBER Force Fields Improves the Description of Z- and B-DNA

Z-DNA duplexes are a particularly complicated test case for current force fields. We performed a set of explicit solvent molecular dynamics (MD) simulations with various AMBER force field parametrizations including our recent refinements of the ε/ζ and glycosidic torsions. None of these force fields described the ZI/ZII and other backbone substates correctly, and all of them underpredicted the population of the important ZI substate. We show that this underprediction can be attributed to an inaccurate potential for the sugar-phosphate backbone torsion angle β. We suggest a refinement of this potential, β(OL1), which was derived using our recently introduced methodology that includes conformation-dependent solvation effects. The new potential significantly increases the stability of the dominant ZI backbone substate and improves the overall description of the Z-DNA backbone. It also has a positive (albeit small) impact on another important DNA form, the antiparallel guanine quadruplex (G-DNA), and improves the description of the canonical B-DNA backbone by increasing the population of BII backbone substates, providing a better agreement with experiment. We recommend using β(OL1) in combination with our previously introduced corrections, εζ(OL1) and χ(OL4), (the combination being named OL15) as a possible alternative to the current β torsion potential for more accurate modeling of nucleic acids.

Protein flexibility in the light of structural alphabets

Protein structures are valuable tools to understand protein function. Nonetheless, proteins are often considered as rigid macromolecules while their structures exhibit specific flexibility, which is essential to complete their functions. Analyses of protein structures and dynamics are often performed with a simplified three-state description, i.e., the classical secondary structures. More precise and complete description of protein backbone conformation can be obtained using libraries of small protein fragments that are able to approximate every part of protein structures. These libraries, called structural alphabets (SAs), have been widely used in structure analysis field, from definition of ligand binding sites to superimposition of protein structures. SAs are also well suited to analyze the dynamics of protein structures. Here, we review innovative approaches that investigate protein flexibility based on SAs description. Coupled to various sources of experimental data (e.g., B-factor) and computational methodology (e.g., Molecular Dynamic simulation), SAs turn out to be powerful tools to analyze protein dynamics, e.g., to examine allosteric mechanisms in large set of structures in complexes, to identify order/disorder transition. SAs were also shown to be quite efficient to predict protein flexibility from amino-acid sequence. Finally, in this review, we exemplify the interest of SAs for studying flexibility with different cases of proteins implicated in pathologies and diseases.

Bioinformatic analysis of the protein/DNA interface

To investigate the principles driving recognition between proteins and DNA, we analyzed more than thousand crystal structures of protein/DNA complexes. We classified protein and DNA conformations by structural alphabets, protein blocks [de Brevern, Etchebest and Hazout (2000) (Bayesian probabilistic approach for predicting backbone structures in terms of protein blocks. Prots. Struct. Funct. Genet., 41:271-287)] and dinucleotide conformers [Svozil, Kalina, Omelka and Schneider (2008) (DNA conformations and their sequence preferences. Nucleic Acids Res., 36:3690-3706)], respectively. Assembling the mutually interacting protein blocks and dinucleotide conformers into 'interaction matrices' revealed their correlations and conformer preferences at the interface relative to their occurrence outside the interface. The analyzed data demonstrated important differences between complexes of various types of proteins such as transcription factors and nucleases, distinct interaction patterns for the DNA minor groove relative to the major groove and phosphate and importance of water-mediated contacts. Water molecules mediate proportionally the largest number of contacts in the minor groove and form the largest proportion of contacts in complexes of transcription factors. The generally known induction of A-DNA forms by complexation was more accurately attributed to A-like and intermediate A/B conformers rare in naked DNA molecules.