Δ-Quantum machine learning for medicinal chemistry

Many molecular design tasks benefit from fast and accurate calculations of quantum-mechanical (QM) properties. However, the computational cost of QM methods applied to drug-like molecules currently renders large-scale applications of quantum chemistry challenging. Aiming to mitigate this problem, we developed DelFTa, an open-source toolbox for the prediction of electronic properties of drug-like molecules at the density functional (DFT) level of theory, using Δ-machine-learning. Δ-Learning corrects the prediction error (Δ) of a fast but inaccurate property calculation. DelFTa employs state-of-the-art three-dimensional message-passing neural networks trained on a large dataset of QM properties. It provides access to a wide array of quantum observables on the molecular, atomic and bond levels by predicting approximations to DFT values from a low-cost semiempirical baseline. Δ-Learning outperformed its direct-learning counterpart for most of the considered QM endpoints. The results suggest that predictions for non-covalent intra- and intermolecular interactions can be extrapolated to larger biomolecular systems. The software is fully open-sourced and features documented command-line and Python APIs.

Optimal spatial allocation of enzymes as an investment problem

Given a limited number of molecular components, cells face various allocation problems demanding decisions on how to distribute their resources. For instance, cells decide which enzymes to produce at what quantity, but also where to position them. Here we focus on the spatial allocation problem of how to distribute enzymes such as to maximize the total reaction flux produced by them in a system with given geometry and boundary conditions. So far, such distributions have been studied by computational optimization, but a deeper theoretical understanding was lacking. We derive an optimal allocation principle, which demands that the available enzymes are distributed such that the marginal flux returns at each occupied position are equal. This ‘homogeneous marginal returns criterion’ (HMR criterion) corresponds to a portfolio optimization criterion in a scenario where each investment globally feeds back onto all payoffs. The HMR criterion allows us to analytically understand and characterize a localization-delocalization transition in the optimal enzyme distribution that was previously observed numerically. In particular, our analysis reveals the generality of the transition, and produces a practical test for the optimality of enzyme localization by comparing the reaction flux to the influx of substrate. Based on these results, we devise an additive construction algorithm, which builds up optimal enzyme arrangements systematically rather than by trial and error. Taken together, our results reveal a common principle in allocation problems from biology and economics, which can also serve as a design principle for synthetic biomolecular systems.

Collective residue interactions in trimer complexes of SARS-CoV-2 spike proteins analyzed by fragment molecular orbital method

In large biomolecular systems such as protein complexes, there are huge numbers of combinations of inter-residue interactions whose comprehensive analyses are often beyond the intuitive processing by researchers. Here we propose a computational method to allow for a systematic analysis of these interactions based on the fragment molecular orbital calculations, in which the inter-fragment interaction energies are comprehensively processed by the singular value decomposition. For a trimer complex of SARS-CoV-2 spike protein, three-body interactions among residues belonging to three chains are analyzed to elicit a small number of essential interaction modes or networks crucial for the structural stability of complex.

Robust finite automata in stochastic chemical reaction networks

Finite-state automata (FSA) are simple computational devices that can nevertheless illustrate interesting behaviours. We propose that FSA can be employed as control circuits for engineered stochastic biological and biomolecular systems. We present an implementation of FSA using counts of chemical species in the range of hundreds to thousands, which is relevant for the counts of many key molecules such as mRNAs in prokaryotic cells. The challenge here is to ensure a robust representation of the current state in the face of stochastic noise. We achieve this by using a multistable approximate majority algorithm to stabilize and store the current state of the system. Arbitrary finite state machines can thus be compiled into robust stochastic chemical automata. We present two variants: one that consumes its input signals to initiate state transitions and one that does not. We characterize the state change dynamics of these systems and demonstrate their application to solve the four-bit binary square root problem. Our work lays the foundation for the use of chemical automata as control circuits in bioengineered systems and biorobotics.

Critical role of backbone coordination in the mRNA recognition by RNA induced silencing complex

Despite its functional importance, the molecular mechanism underlying target mRNA recognition by Argonaute (Ago) remains largely elusive. Based on extensive all-atom molecular dynamics simulations, we constructed quasi-Markov State Model (qMSM) to reveal the dynamics during recognition at position 6-7 in the seed region of human Argonaute 2 (hAgo2). Interestingly, we found that the slowest mode of motion therein is not the gRNA-target base-pairing, but the coordination of the target phosphate groups with a set of positively charged residues of hAgo2. Moreover, the ability of Helix-7 to approach the PIWI and MID domains was found to reduce the effective volume accessible to the target mRNA and therefore facilitate both the backbone coordination and base-pair formation. Further mutant simulations revealed that alanine mutation of the D358 residue on Helix-7 enhanced a trap state to slow down the loading of target mRNA. Similar trap state was also observed when wobble pairs were introduced in g6 and g7, indicating the role of Helix-7 in suppressing non-canonical base-paring. Our study pointed to a general mechanism for mRNA recognition by eukaryotic Agos and demonstrated the promise of qMSM in investigating complex conformational changes of biomolecular systems.

Expanding the Current Boundaries of Nature-based Modeling and Computing: Chem-Inspiration for Meta-Heuristics

High-throughput methods in science have created a trend to generate massive amount of data that challenge our ability to mine and search through massive information spaces. Thus more efficient and effective solutions for data analysis and optimization are required continuously. The best solutions for many problems-solving approaches in science could have many sources of inspiration coming from diverse natural phenomena. In this context, most Artificial Intelligence (AI) approaches benefit from emulation natural processes for their information processing strategy. Among the AI protocols, meta-heuristic algorithms for learning model and optimization have exploited a number of biological phenomena leading to highly effective search and learning engines. Examples of these processes are the ant colony organization, brain function and genetics among others. The evolution has turned all these biological events in highly efficient procedures, whose basics principles have then provided an excellent ground of new computational algorithms The aim of this report is pave the way to a new class of nature-based meta-heuristic methods which shall be based on diverse chemical and biomolecular systems. We present five examples from different subjects of Chemistry like Organic Chemistry, Chemical Physics and Biomolecules; and introduce how computational models could be inferred from them. Besides, we develop one of these models, in detail, which is based on protein evolution and folding principles. We consider that the wealth of systems and processes related to Chemistry, as those described in the present communication, might boost the development of relevant meta-heuristic and classification algorithms in upcoming years.

Improvement of the Force Field for β-d-Glucose with Machine Learning

While the construction of a dependable force field for performing classical molecular dynamics (MD) simulation is crucial for elucidating the structure and function of biomolecular systems, the attempts to do this for glycans are relatively sparse compared to those for proteins and nucleic acids. Currently, the use of GLYCAM06 force field is the most popular, but there have been a number of concerns about its accuracy in the systematic description of structural changes. In the present work, we focus on the improvement of the GLYCAM06 force field for β-d-glucose, a simple and the most abundant monosaccharide molecule, with the aid of machine learning techniques implemented with the TensorFlow library. Following the pre-sampling over a wide range of configuration space generated by MD simulation, the atomic charge and dihedral angle parameters in the GLYCAM06 force field were re-optimized to accurately reproduce the relative energies of β-d-glucose obtained by the density functional theory (DFT) calculations according to the structural changes. The validation for the newly proposed force-field parameters was then carried out by verifying that the relative energy errors compared to the DFT value were significantly reduced and that some inconsistencies with experimental (e.g., NMR) results observed in the GLYCAM06 force field were resolved relevantly.

Post-translational insertion of boron in proteins to probe and modulate function

Boron is absent in proteins, yet is a micronutrient. It possesses unique bonding that could expand biological function including modes of Lewis acidity not available to typical elements of life. Here we show that post-translational Cβ–Bγ bond formation provides mild, direct, site-selective access to the minimally sized residue boronoalanine (Bal) in proteins. Precise anchoring of boron within complex biomolecular systems allows dative bond-mediated, site-dependent protein Lewis acid–base-pairing (LABP) by Bal. Dynamic protein-LABP creates tunable inter- and intramolecular ligand–host interactions, while reactive protein-LABP reveals reactively accessible sites through migratory boron-to-oxygen Cβ–Oγ covalent bond formation. These modes of dative bonding can also generate de novo function, such as control of thermo- and proteolytic stability in a target protein, or observation of transient structural features via chemical exchange. These results indicate that controlled insertion of boron facilitates stability modulation, structure determination, de novo binding activities and redox-responsive ‘mutation’.

An automated framework for high-throughput predictions of NMR chemical shifts within liquid solutions

Identifying stable speciation in multicomponent liquid solutions is of fundamental importance to areas ranging from electrochemistry to organic chemistry and biomolecular systems. However, elucidating this complex solvation environment is a daunting task even when using advanced experimental and computational techniques. Here, we introduce a fully automated, high-throughput computational framework for the accurate and robust prediction of stable species present in liquid solutions by computing the nuclear magnetic resonance (NMR) chemical shifts of molecules. The framework automatically extracts and categorizes hundreds of thousands of atomic clusters from classical molecular dynamics (CMD) simulations to identify the most stable speciation in the solution and calculate their NMR chemical shifts via DFT calculations. Additionally, the framework creates an output database of computed chemical shifts for liquid solutions across a wide chemical and parameter space. This task can be infeasible experimentally and challenging using conventional computational methods. To demonstrate the capabilities of our framework, we compare our computational results to experimental measurements for a complex test case of magnesium bis(trifluoromethanesulfonyl)imide Mg(TFSI)2 salt in dimethoxyethane (DME) solvent, which is a common electrolyte system for Mg-based batteries. Our extensive benchmarking and analysis of the Mg2+ solvation structural evolutions reveal critical factors such as the effect of force field parameters that influence the accuracy of NMR chemical shift predictions in liquid solutions. Furthermore, we show how the framework reduces the efforts of performing and managing over 300 13C and 600 1H DFT chemical shift predictions to a single submission procedure. By enabling more efficient and accurate high-throughput computations of NMR chemical shifts, our approach can accelerate theory-guided design of liquid solutions for various applications.