Machine learning models for hydrogen bond donor and acceptor strengths using large and diverse training data generated by first-principles interaction free energies

Synergy of machine learning and density functional theory calculations for predicting experimental Lewis base affinity and Lewis polybase binding atoms

Investigation of Lewis acid-base interactions has been conducted by ab initio calculations and machine learning (ML) models. This study aims to resolve two critical tasks that have not been quantitatively investigated. First, ML models developed from density functional theory (DFT) calculations predict experimental BF3 affinity with Pearson correlation coefficients around 0.9 and mean absolute errors around 10 kJ mol-1. The ML models are trained by DFT-calculated BF3 affinity of more than 3000 adducts, with input features readily obtained by rdkit. Second, the ML models have the capability of predicting the relative strength of Lewis base binding atoms in Lewis polybases, which is either an extremely challenging task to conduct experimentally or a computationally expensive task for ab initio methods. The study demonstrates and solidifies the potential of combining DFT calculations and ML models to predict experimental properties, especially those that are scarce and impractical to empirically acquire.

Design, synthesis, electrochemistry and anti-trypanosomatid hit/lead identification of nitrofuranylazines

Chagas disease and leishmaniasis are vector-borne infectious diseases affecting both humans and animals. These neglected tropical diseases can be fatal if not treated. Hundreds to thousands of new Chagas disease and leishmaniasis cases are being reported by the WHO every year, and currently available treatments are insufficient. Severe adverse effects, impractical administrations and increased pathogen resistance against current clinical treatments underscore a serious need for the development of new drugs to curb these ailments. In search for such drugs, we investigated a series of nitrofuran-based azine derivatives. Herein, we report the design, synthesis, electrochemistry, and biological activity of these derivatives against promastigotes and amastigotes of Leishmania major, and L. donovani strains, as well as epimastigotes and trypomastigotes of Trypanosoma cruzi. Two leishmanicidal early leads and one trypanosomacidal hit with submicromolar activity were uncovered and stand for further in vivo investigation in the search for new antitrypanosomatid drugs. Future objective will focus on the identification of involved biological targets with the parasites.

Fast calculation of hydrogen-bond strengths and free energy of hydration of small molecules

Hydrogen bonding is an interaction of great importance in drug discovery and development as it may significantly affect chemical and biological processes including the interaction of small molecules with other molecules, proteins, and membranes. In particular, hydrogen bonding can impact drug-like properties such as target affinity and oral availability which are critical to developing effective pharmaceuticals, and therefore, numerous methods for the calculation of properties such as hydrogen-bond strengths, free energy of hydration, or water solubility have been proposed over time. However, the accessibility to efficient methods for the predictions of such properties is still limited. Here, we present the development of Jazzy, an open-source tool for the prediction of hydrogen-bond strengths and free energies of hydration of small molecules. Jazzy also allows the visualisation of hydrogen-bond strengths with atomistic resolution to support the design of compounds with desired properties and the interpretation of existing data. The tool is described in its implementation, parameter fitting, and validation against two data sets of experimental hydration free energies. Jazzy is also applied against two chemical series of bioactive compounds to show that hydrogen-bond strengths can be used to understand their structure-activity relationships. Results from the validations highlight the strengths and limitations of Jazzy, and suggest its suitability for interactive design, screening, and machine-learning featurisation.

Predicting Solute Descriptors for Organic Chemicals by a Deep Neural Network (DNN) Using Basic Chemical Structures and a Surrogate Metric

Solute descriptors have been widely used to model chemical transfer processes through poly-parameter linear free energy relationships (pp-LFERs); however, there are still substantial difficulties in obtaining these descriptors accurately and quickly for new organic chemicals. In this research, models (PaDEL-DNN) that require only SMILES of chemicals were built to satisfactorily estimate pp-LFER descriptors using deep neural networks (DNN) and the PaDEL chemical representation. The PaDEL-DNN-estimated pp-LFER descriptors demonstrated good performance in modeling storage-lipid/water partitioning coefficient (log K_{storage-lipid/water}), bioconcentration factor (BCF), aqueous solubility (ESOL), and hydration free energy (freesolve). Then, assuming that the accuracy in the estimated values of widely available properties, e.g., logP (octanol-water partition coefficient), can calibrate estimates for less available but related properties, we proposed logP as a surrogate metric for evaluating the overall accuracy of the estimated pp-LFER descriptors. When using the pp-LFER descriptors to model log K_{storage-lipid/water}, BCF, ESOL, and freesolve, we achieved around 0.1 log unit lower errors for chemicals whose estimated pp-LFER descriptors were deemed "accurate" by the surrogate metric. The interpretation of the PaDEL-DNN models revealed that, for a given test chemical, having several (around 5) "similar" chemicals in the training data set was crucial for accurate estimation while the remaining less similar training chemicals provided reasonable baseline estimates. Lastly, pp-LFER descriptors for over 2800 persistent, bioaccumulative, and toxic chemicals were reasonably estimated by combining PaDEL-DNN with the surrogate metric. Overall, the PaDEL-DNN/surrogate metric and newly estimated descriptors will greatly benefit chemical transfer modeling.

Modulating physicochemical properties of tetrahydropyridine-2-amine BACE1 inhibitors with electron-withdrawing groups: A systematic study

A common challenge for medicinal chemists is to reduce the pK_a of strongly basic groups' conjugate acids into a range that preserves the desired effects, usually potency and/or solubility, but avoids undesired effects like high volume of distribution (V_d), limited membrane permeation, and off-target binding to, notably, the hERG channel and monoamine receptors. We faced this challenge with a 3,4,5,6-tetrahydropyridine-2-amine scaffold harboring an amidine, a key structural component of potential inhibitors of BACE1, the rate-limiting enzyme in the production of Aβ species that make up amyloid plaques in Alzheimer's disease. In our endeavor to balance potency with desirable properties to achieve brain penetration, we introduced a diverse set of groups in beta position of the amidine that modulate logD, PSA and pK_a. Given the synthetic challenge to prepare these highly functionalized warheads, we first developed a design flow including predicted physicochemical parameters which allowed us to select only the most promising candidates for synthesis. For this we evaluated a set of commercial packages to predict physicochemical properties, which can guide medicinal chemists in their endeavors to modulate pK_a values of amidine and amine bases.

Chemical reactivity prediction: current methods and different application areas

The ability to predict chemical reactivity of a molecule is highly desirable in drug discovery, both ex vivo (synthetic route planning, formulation, stability) and in vivo: metabolic reactions determine pharmacodynamics, pharmacokinetics and potential toxic effects, and early assessment of liabilities is vital to reduce attrition rates in later stages of development. Quantum mechanics offer a precise description of the interactions between electrons and orbitals in the breaking and forming of new bonds. Modern algorithms and faster computers have allowed the study of more complex systems in a punctual and accurate fashion, and answers for chemical questions around stability and reactivity can now be provided. Through machine learning, predictive models can be built out of descriptors derived from quantum mechanics and cheminformatics, even in the absence of experimental data to train on. In this article, current progress on computational reactivity prediction is reviewed: applications to problems in drug design, such as modelling of metabolism and covalent inhibition, are highlighted and unmet challenges are posed.

Machine Learning Applied to the Modeling of Pharmacological and ADMET Endpoints

The well-known concept of quantitative structure-activity relationships (QSAR) has been gaining significant interest in the recent years. Data, descriptors, and algorithms are the main pillars to build useful models that support more efficient drug discovery processes with in silico methods. Significant advances in all three areas are the reason for the regained interest in these models. In this book chapter we review various machine learning (ML) approaches that make use of measured in vitro/in vivo data of many compounds. We put these in context with other digital drug discovery methods and present some application examples.

Prediction of Chameleonic Efficiency

Chameleonic properties, i. e., the capacity of a molecule to hide polarity in non-polar environments and expose it in water, help achieving sufficient permeability and solubility for drug molecules with high MW. We present models of experimental measures of polarity for a set of 24 FDA approved drugs (MW 405-1113) and one PROTAC (MW 1034). Conformational ensembles in aqueous and non-polar environments were generated using molecular dynamics. A linear regression model that predicts chromatographic apparent polarity (EPSA) with a mean unsigned error of 10 Ų was derived based on separate terms for donor, acceptor, and total molecular SASA. A good correlation (R² =0.92) with an experimental measure of hydrogen bond donor potential, Δlog P_oct-tol , was found for the mean hydrogen bond donor SASA of the conformational ensemble scaled with Abraham's A hydrogen bond acidity. Two quantitative measures of chameleonic behaviour, the chameleonic efficiency indices, are introduced. We envision that the methods presented herein will be useful to triage designed molecules and prioritize those with the best chance of achieving acceptable permeability and solubility.

Artificial intelligence in drug design: algorithms, applications, challenges and ethics

The discovery paradigm of drugs is rapidly growing due to advances in machine learning (ML) and artificial intelligence (AI). This review covers myriad faces of AI and ML in drug design. There is a plethora of AI algorithms, the most common of which are summarized in this review. In addition, AI is fraught with challenges that are highlighted along with plausible solutions to them. Examples are provided to illustrate the use of AI and ML in drug discovery and in predicting drug properties such as binding affinities and interactions, solubility, toxicology, blood–brain barrier permeability and chemical properties. The review also includes examples depicting the implementation of AI and ML in tackling intractable diseases such as COVID-19, cancer and Alzheimer’s disease. Ethical considerations and future perspectives of AI are also covered in this review.

Bayer's in silico ADMET platform: a journey of machine learning over the past two decades

Over the past two decades, an in silico absorption, distribution, metabolism, and excretion (ADMET) platform has been created at Bayer Pharma with the goal to generate models for a variety of pharmacokinetic and physicochemical endpoints in early drug discovery. These tools are accessible to all scientists within the company and can be a useful in assisting with the selection and design of novel leads, as well as the process of lead optimization. Here. we discuss the development of machine-learning (ML) approaches with special emphasis on data, descriptors, and algorithms. We show that high company internal data quality and tailored descriptors, as well as a thorough understanding of the experimental endpoints, are essential to the utility of our models. We discuss the recent impact of deep neural networks and show selected application examples.

Approximating the Strength of the Intramolecular Hydrogen Bond in 2-Fluorophenol and Related Compounds: A New Application of a Classic Technique

Fluorinated organic compounds are ubiquitous in the pharmaceutical and agricultural industries. To better discern the mode of action of these compounds, it is critical to understand the strengths of hydrogen bonds involving fluorine. While established techniques can determine these strengths for intermolecular complexes, there is no analogous scheme for intramolecular hydrogen bonds. This work uses ¹H nuclear magnetic resonance spectroscopy to measure the strength of intramolecular hydrogen bonds in ortho-substituted phenols. Titration of each phenol with DMSO in CCl₄ yields a free energy of binding (ΔG). Subtraction of this value from the ΔG of binding of the standard, 4-fluorophenol, is shown to give the difference in ΔG for the cis and trans isomers of the ortho-substituted phenols. This difference is conventionally taken to be approximately equal to the ΔG of the intramolecular hydrogen bond. These data complement theoretical methods, which yield slightly larger ΔGs. Both theory and experiment point to a weak intramolecular hydrogen bond in 2-fluorophenol. The other 2-X-phenols have stronger hydrogen bonds, following the order F < Cl ≈ Br < OCH₃. The methodology developed here can be readily applied to other systems with intramolecular hydrogen bonds.