Machine Learning Prediction of DNA Charge Transport

Ultrafast excitonic dynamics in DNA: Bridging correlated quantum dynamics and sequence dependence

After photoexcitation of DNA, the excited electron (in the LUMO) and the remaining hole (in the HOMO) localized on the same DNA base form a bound pair, called the Frenkel exciton, due to their mutual Coulomb interaction. In this study, we demonstrate that a tight-binding (TB) approach, using TB parameters for electrons and holes available in the literature, allows us to correlate relaxation properties, average charge separation, and dipole moments to a large ensemble of double-stranded DNA sequences (all 16384 possible sequences with 14 nucleobases). This way, we are able to identify a relatively small subensemble of sequences responsible for long-lived excited states, high average charge separation, and high dipole moment. Further analysis shows that these sequences are particularly T rich. By systematically screening the impact of electron-hole interaction (Coulomb forces), we verify that these correlations are relatively robust against finite-size variations of the interaction parameter, not directly accessible experimentally. This methodology combines simulation methods from quantum physics and physical chemistry with statistical analysis known from genetics and epigenetics, thus representing a powerful bridge to combine information from both fields.

From density functional theory to machine learning predictive models for electrical properties of spinel oxides

This work focuses on predicting and characterizing the electronic conductivity of spinel oxides, which are promising materials for energy storage devices and for the oxygen evolution and oxygen reduction reactions due to their attractive properties and abundance of transition metals that can act as active sites for catalysis. To this end, a new database was developed from first principles, including band structure and conductivity properties of spinel oxides, and machine learning algorithms were trained on this database to predict electronic conductivity and band gaps based solely on the compositions. The models developed in this study are scaled from the quantum level up to a continuum conductivity model. The relatively small database used in this study allowed for accurate predictions of band gap and conductivity. By altering the composition of spinel oxides, the model was able to predict high conductivity for spinels with high nickel content and to match experimental trends for manganese cobalt spinels. The ability to predict material properties is especially important in energy conversion devices such as batteries and supercapacitors where redox reactions take place.

Emergence of Boltzmann Subspaces in Open Quantum Systems Far from Equilibrium

Single molecule junctions are important examples of complex out-of-equilibrium many-body quantum systems. We identify a nontrivial clustering of steady state populations into distinctive subspaces with Boltzmann-like statistics, which persist far from equilibrium. Such Boltzmann subspaces significantly reduce the information needed to describe the steady state, enabling modeling of high-dimensional systems that are otherwise beyond the reach of current computations. The emergence of Boltzmann subspaces is demonstrated analytically and numerically for fermionic transport systems of increasing complexity.

Modeling of H2S solubility in ionic liquids: comparison of white-box machine learning, deep learning and ensemble learning approaches

In the context of gas processing and carbon sequestration, an adequate understanding of the solubility of acid gases in ionic liquids (ILs) under various thermodynamic circumstances is crucial. A poisonous, combustible, and acidic gas that can cause environmental damage is hydrogen sulfide (H₂S). ILs are good choices for appropriate solvents in gas separation procedures. In this work, a variety of machine learning techniques, such as white-box machine learning, deep learning, and ensemble learning, were established to determine the solubility of H₂S in ILs. The white-box models are group method of data handling (GMDH) and genetic programming (GP), the deep learning approach is deep belief network (DBN) and extreme gradient boosting (XGBoost) was selected as an ensemble approach. The models were established utilizing an extensive database with 1516 data points on the H₂S solubility in 37 ILs throughout an extensive pressure and temperature range. Seven input variables, including temperature (T), pressure (P), two critical variables such as temperature (T_c) and pressure (P_c), acentric factor (ω), boiling temperature (T_b), and molecular weight (Mw), were used in these models; the output was the solubility of H₂S. The findings show that the XGBoost model, with statistical parameters such as an average absolute percent relative error (AAPRE) of 1.14%, root mean square error (RMSE) of 0.002, standard deviation (SD) of 0.01, and a determination coefficient (R²) of 0.99, provides more precise calculations for H₂S solubility in ILs. The sensitivity assessment demonstrated that temperature and pressure had the highest negative and highest positive affect on the H₂S solubility in ILs, respectively. The Taylor diagram, cumulative frequency plot, cross-plot, and error bar all demonstrated the high effectiveness, accuracy, and reality of the XGBoost approach for predicting the H₂S solubility in various ILs. The leverage analysis shows that the majority of the data points are experimentally reliable and just a small number of data points are found beyond the application domain of the XGBoost paradigm. Beyond these statistical results, some chemical structure effects were evaluated. First, it was shown that the lengthening of the cation alkyl chain enhances the H₂S solubility in ILs. As another chemical structure effect, it was shown that higher fluorine content in anion leads to higher solubility in ILs. These phenomena were confirmed by experimental data and the model results. Connecting solubility data to the chemical structure of ILs, the results of this study can further assist to find appropriate ILs for specialized processes (based on the process conditions) as solvents for H₂S.

Nucleic Acid Quantification by Multi-Frequency Impedance Cytometry and Machine Learning

Determining nucleic acid concentrations in a sample is an important step prior to proceeding with downstream analysis in molecular diagnostics. Given the need for testing DNA amounts and its purity in many samples, including in samples with very small input DNA, there is utility of novel machine learning approaches for accurate and high-throughput DNA quantification. Here, we demonstrated the ability of a neural network to predict DNA amounts coupled to paramagnetic beads. To this end, a custom-made microfluidic chip is applied to detect DNA molecules bound to beads by measuring the impedance peak response (IPR) at multiple frequencies. We leveraged electrical measurements including the frequency and imaginary and real parts of the peak intensity within a microfluidic channel as the input of deep learning models to predict DNA concentration. Specifically, 10 different deep learning architectures are examined. The results of the proposed regression model indicate that an R_Squared of 97% with a slope of 0.68 is achievable. Consequently, machine learning models can be a suitable, fast, and accurate method to measure nucleic acid concentration in a sample. The results presented in this study demonstrate the ability of the proposed neural network to use the information embedded in raw impedance data to predict the amount of DNA concentration.

Learning Conductance: Gaussian Process Regression for Molecular Electronics

Experimental studies of charge transport through single molecules often rely on break junction setups, where molecular junctions are repeatedly formed and broken while measuring the conductance, leading to a statistical distribution of conductance values. Modeling this experimental situation and the resulting conductance histograms is challenging for theoretical methods, as computations need to capture structural changes in experiments, including the statistics of junction formation and rupture. This type of extensive structural sampling implies that even when evaluating conductance from computationally efficient electronic structure methods, which typically are of reduced accuracy, the evaluation of conductance histograms is too expensive to be a routine task. Highly accurate quantum transport computations are only computationally feasible for a few selected conformations and thus necessarily ignore the rich conformational space probed in experiments. To overcome these limitations, we investigate the potential of machine learning for modeling conductance histograms, in particular by Gaussian process regression. We show that by selecting specific structural parameters as features, Gaussian process regression can be used to efficiently predict the zero-bias conductance from molecular structures, reducing the computational cost of simulating conductance histograms by an order of magnitude. This enables the efficient calculation of conductance histograms even on the basis of computationally expensive first-principles approaches by effectively reducing the number of necessary charge transport calculations, paving the way toward their routine evaluation.

Long-range charge transport in homogeneous and alternating-rigidity chains

We study the interplay of intrinsic-electronic and environmental factors in long-range charge transport across molecular chains with up to N ∼ 80 monomers. We describe the molecular electronic structure of the chain with a tight-binding Hamiltonian. Thermal effects in the form of electron decoherence and inelastic scattering are incorporated with the Landauer–Büttiker probe method. In short chains of up to ten units, we observe the crossover between coherent (tunneling, ballistic) motion and thermally-assisted conduction, with thermal effects enhancing the current beyond the quantum coherent limit. We further show that unconventional (nonmonotonic with size) transport behavior emerges when monomer-to-monomer electronic coupling is made large. In long chains, we identify a different behavior, with thermal effects suppressing the conductance below the coherent-ballistic limit. With the goal to identify a minimal model for molecular chains displaying unconventional and effective long-range transport, we simulate a modular polymer with alternating regions of high and low rigidity. Simulations show that, surprisingly, while charge correlations are significantly affected by structuring environmental conditions, reflecting charge delocalization, the electrical resistance displays an averaging effect, and it is not sensitive to this patterning. We conclude by arguing that efficient long-range charge transport requires engineering both internal electronic parameters and environmental conditions.

Deciphering quantum fingerprints in electric conductance

When the electric conductance of a nano-sized metal is measured at low temperatures, it often exhibits complex but reproducible patterns as a function of external magnetic fields called quantum fingerprints in electric conductance. Such complex patterns are due to quantum–mechanical interference of conduction electrons; when thermal disturbance is feeble and coherence of the electrons extends all over the sample, the quantum interference pattern reflects microscopic structures, such as crystalline defects and the shape of the sample, giving rise to complicated interference. Although the interference pattern carries such microscopic information, it looks so random that it has not been analysed. Here we show that machine learning allows us to decipher quantum fingerprints; fingerprint patterns in magneto-conductance are shown to be transcribed into spatial images of electron wave function intensities (WIs) in a sample by using generative machine learning. The output WIs reveal quantum interference states of conduction electrons, as well as sample shapes. The present result augments the human ability to identify quantum states, and it should allow microscopy of quantum nanostructures in materials by making use of quantum fingerprints.

Scattering of electrons from defects and boundaries in mesoscopic samples is encoded in quantum interference patterns of magneto-conductance, but these patterns are difficult to interpret. Here the authors use machine learning to reconstruct electron wavefunction intensities and sample geometry from magneto-conductance data.

Using Weakly Supervised Deep Learning to Classify and Segment Single-Molecule Break-Junction Conductance Traces

In order to design molecular electronic devices with high performance and stability, it is crucial to understand their structure-to-property relationships. Single-molecule break junction measurements yield a large number of conductance-distance traces, which are inherently highly stochastic. Here we propose a weakly supervised deep learning algorithm to classify and segment these conductance traces, a method that is mainly based on transfer learning with the pretrain-finetune technique. By exploiting the powerful feature extraction capabilities of the pretrained VGG-16 network, our convolutional neural network model not only achieves high accuracy in the classification of the conductance traces, but also segments precisely the conductance plateau from an entire trace with very few manually labeled traces. Thus, we can produce a more reliable estimation of the junction conductance and quantify the junction stability. These findings show that our model has achieved a better accuracy-to-manpower efficiency balance, opening up the possibility of using weakly supervised deep learning approaches in the studies of single-molecule junctions.

Deep-Learning Approach to First-Principles Transport Simulations

Large-scale first-principles transport calculations, while essential for device modeling, remain computationally demanding. To overcome this bottle neck, we combine first-principles transport calculations with machine learning-based nonlinear regression. We calculate the electronic conductance through first-principles based nonequilibrium Green's function techniques for small systems and map the transport properties onto local properties using local descriptors. We show that using the local descriptor as input features for deep learning-based nonlinear regression allows us to build a robust neural network that can predict the conductance of large systems beyond that of the current state-of-the-art first-principles calculation algorithms. Our protocol is applied to alkali metal nanowires, i.e., potassium, which have unique geometrical and electronic properties and hence nontrivial transport properties. We demonstrate that within our approach we can achieve qualitative agreement with experiment at a fraction of the computational effort as compared to the direct calculation of the transport properties using conventional first-principles methods.

Predicting the DNA Conductance Using a Deep Feedforward Neural Network Model

Double-stranded DNA (dsDNA) has been established as an efficient medium for charge migration, bringing it to the forefront of the field of molecular electronics and biological research. The charge migration rate is controlled by the electronic couplings between the two nucleobases of DNA/RNA. These electronic couplings strongly depend on the intermolecular geometry and orientation. Estimating these electronic couplings for all the possible relative geometries of molecules using the computationally demanding first-principles calculations requires a lot of time and computational resources. In this article, we present a machine learning (ML)-based model to calculate the electronic coupling between any two bases of dsDNA/dsRNA and bypass the computationally expensive first-principles calculations. Using the Coulomb matrix representation which encodes the atomic identities and coordinates of the DNA base pairs to prepare the input dataset, we train a feedforward neural network model. Our neural network (NN) model can predict the electronic couplings between dsDNA base pairs with any structural orientation with a mean absolute error (MAE) of less than 0.014 eV. We further use the NN-predicted electronic coupling values to compute the dsDNA/dsRNA conductance.

Toward a First-Principles Evaluation of Transport Mechanisms in Molecular Wires

Understanding charge transport through molecular wires is important for nanoscale electronics and biochemistry. Our goal is to establish a simple first-principles protocol for predicting the charge transport mechanism in such wires, in particular the crossover from coherent tunneling for short wires to incoherent hopping for longer wires. This protocol is based on a combination of density functional theory with a polarizable continuum model introduced by Kaupp et al. for mixed-valence molecules, which we had previously found to work well for length-dependent charge delocalization in such systems. We combine this protocol with a new charge delocalization measure tailored for molecular wires, and we show that it can predict the tunneling-to-hopping transition length with a maximum error of one subunit in five sets of molecular wires studied experimentally in molecular junctions at room temperature. This suggests that the protocol is also well suited for estimating the extent of hopping sites as relevant, for example, for the intermediate tunneling-hopping regime in DNA.

Using Deep Learning to Identify Molecular Junction Characteristics

The scanning tunneling microscope-based break junction (STM-BJ) is used widely to create and characterize single metal-molecule-metal junctions. In this technique, conductance is continuously recorded as a metal point contact is broken in a solution of molecules. Conductance plateaus are seen when stable molecular junctions are formed. Typically, thousands of junctions are created and measured, yielding thousands of distinct conductance versus extension traces. However, such traces are rarely analyzed individually to recognize the types of junctions formed. Here, we present a deep learning-based method to identify molecular junctions and show that it performs better than several commonly used and recently reported techniques. We demonstrate molecular junction identification from mixed solution measurements with accuracies as high as 97%. We also apply this model to an in situ electric field-driven isomerization reaction of a [3]cumulene to follow the reaction over time. Furthermore, we demonstrate that our model can remain accurate even when a key parameter, the average junction conductance, is eliminated from the analysis, showing that our model goes beyond conventional analysis in existing methods.

2'-Deoxy-2'-fluoro-arabinonucleic acid: a valid alternative to DNA for biotechnological applications using charge transport

The non-biological 2'-deoxy-2'-fluoro-arabinonucleic acid (2'F-ANA) may be used as a valid alternative to DNA in biomedical and electronic applications because of its higher resistance to hydrolysis and nuclease degradation. However, the advantage of using 2'F-ANA in such applications also depends on its charge-transfer properties compared to DNA. In this study, we compare the charge conduction properties of model 2'F-ANA and DNA double-strands, using structural snapshots from MD simulations to calculate the electronic couplings and reorganization energies associated with the hole transfer steps between adjacent nucleobase pairs. Inserting these charge-transfer parameters into a kinetic model for charge conduction, we find similar conductive properties for DNA and 2'F-ANA. Moreover, we find that 2'F-ANA's enhanced chemical stability does not correspond to a reduction in the nucleobase π-stack structural flexibility relevant to both electronic couplings and reorganization free energies. Our results promote the use of 2'F-ANA in applications that can be based on charge transport, such as biosensing and chip technology, where its chemical stability and conductivity can advantageously combine.