Deep Learning Models Compared to Experimental Variability for the Prediction of CYP3A4 Time-Dependent Inhibition.

Most drugs are mainly metabolized by cytochrome P450 (CYP450), which can lead to drug-drug interactions (DDI). Specifically, time-dependent inhibition (TDI) of CYP3A4 isoenzyme has been associated with clinically relevant DDI. To overcome potential DDI issues, high-throughput in vitro assays were established to assess the TDI of CYP3A4 during the discovery and lead optimization phases. However, in silico machine learning models would enable an earlier and larger-scale assessment of TDI potential liabilities. For CYP inhibition, most modeling efforts have focused on highly imbalanced and small data sets. Moreover, assay variability is rarely considered, which is key to understand the model's quality and suitability for decision-making. In this work, machine learning models were built for the prediction of TDI of CYP3A4, evaluated prospectively, and compared to the variability of the experimental assay. Different modeling strategies were investigated to assess their influence on the model's performance. Through multitask learning, additional data sets were leveraged for model building, coming from public databases, in-house CYP-related assays, or other pharmaceutical companies (federated learning). Apart from the numerical prediction of inactivation rates of CYP3A4 TDI, three-class predictions were carried out, giving a negative (inactivation rate kobs < 0.01 min-1), weak positive (0.01 ≤ kobs ≤ 0.025 min-1), or positive (kobs > 0.025 min-1) output. The final multitask graph neural network model achieved misclassification rates of 8 and 7% for positive and negative TDI, respectively. Importantly, the presented deep learning-based predictions had a similar precision to the reproducibility of in vitro experiments and thus offered great opportunities for drug design, early derisk of DDI potential, and selection of experiments. To facilitate CYP inhibition modeling efforts in the public domain, the developed model was used to annotate ∼16 000 publicly available structures, and a surrogate data set is shared as Supporting Information.

Adapting Deep Learning QSPR Models to Specific Drug Discovery Projects.

Medicinal chemistry and drug design efforts can be assisted by machine learning (ML) models that relate the molecular structure to compound properties. Such quantitative structure-property relationship models are generally trained on large data sets that include diverse chemical series (global models). In the pharmaceutical industry, these ML global models are available across discovery projects as an "out-of-the-box" solution to assist in drug design, synthesis prioritization, and experiment selection. However, drug discovery projects typically focus on confined parts of the chemical space (e.g., chemical series), where global models might not be applicable. Local ML models are sometimes generated to focus on specific projects or series. Herein, ML-based global models, local models, and hybrid global-local strategies were benchmarked. Analyses were done for more than 300 drug discovery projects at Novartis and ten absorption, distribution, metabolism, and excretion (ADME) assays. In this work, hybrid global-local strategies based on transfer learning approaches were proposed to leverage both historical ADME data (global) and project-specific data (local) to adapt model predictions. Fine-tuning a pretrained global ML model (used for weights' initialization, WI) was the top-performing method. Average improvements of mean absolute errors across all assays were 16% and 27% compared with global and local models, respectively. Interestingly, when the effect of training set size was analyzed, WI fine-tuning was found to be successful even in low-data scenarios (e.g., ∼10 molecules per project). Taken together, this work highlights the potential of domain adaptation in the field of molecular property predictions to refine existing pretrained models on a new compound data distribution.

Systematic Evaluation of Local and Global Machine Learning Models for the Prediction of ADME Properties.

Machine learning (ML) has become an indispensable tool to predict absorption, distribution, metabolism, and excretion (ADME) properties in pharmaceutical research. ML algorithms are trained on molecular structures and corresponding ADME assay data to develop quantitative structure-property relationship (QSPR) models. Traditional QSPR models were trained on compound sets of limited size. With the advent of more complex ML algorithms and data availability, training sets have become larger and more diverse. Most common training approaches consist in either training a model with a small set of similar compounds, namely, compounds designed for the same drug discovery project or chemical series (local model approach) or with a larger set of diverse compounds (global model approach). Global models are built with all experimental data available for an assay, combining compound data from different projects and disease areas. Despite the ML progress made so far, the choice of the appropriate data composition for building ML models is still unclear. Herein, a systematic evaluation of local and global ML models was performed for 10 different experimental assays and 112 drug discovery projects. Results show a consistent superior performance of global models for ADME property predictions. Diagnostic analyses were also carried out to investigate the influence of training set size, structural diversity, and data shift in the relative performance of local and global ML models. Training set and structural diversity did not have an impact in the relative performance on the methods. Instead, data shift helped to identify the projects with larger performance differences between local and global models. Results presented in this work can be leveraged to improve ML-based ADME properties predictions and thus decision-making in drug discovery projects.

Multispecies Machine Learning Predictions of In Vitro Intrinsic Clearance with Uncertainty Quantification Analyses.

In pharmaceutical research, compounds are optimized for metabolic stability to avoid a too fast elimination of the drug. Intrinsic clearance (CLint) measured in liver microsomes or hepatocytes is an important parameter during lead optimization. In this work, machine learning models were developed to relate the compound structure to microsomal metabolic stability and predict CLint for new compounds. A multitask (MT) learning architecture was introduced to model the CLint of six species simultaneously, giving as a result a multispecies machine learning model. MT graph neural network (MT-GNN) regression was identified as the top-performing method, and an ensemble of 10 MT-GNN models was evaluated prospectively. Geometric mean fold errors were consistently smaller than 2-fold. Moreover, high precision values were obtained in the prediction of "high" (>300 µL/min/mg) and "low" (<100 µL/min/mg) CLint compounds. Precision values ranged from 80 to 94% for low CLint predictions and from 75 to 97% for high CLint predictions, depending on the species. Uncertainty on experimental values and model predictions was systematically quantified. Experimental variability (aleatoric uncertainty) of all historical Novartis in vitro clearance experiments was analyzed. Interestingly, MT-GNN models' performance approached assays' experimental variability. Moreover, uncertainty estimation in predictions (epistemic uncertainty) enabled identifying predictions associated with lower and higher error. Taken together, our manuscript combines a multispecies deep learning model and large-scale uncertainty analyses to improve CLint predictions and facilitate early informed decisions for compound prioritization.

Predicting In Vivo Compound Brain Penetration Using Multi-task Graph Neural Networks.

Assessing whether compounds penetrate the brain can become critical in drug discovery, either to prevent adverse events or to reach the biological target. Generally, pre-clinical in vivo studies measuring the ratio of brain and blood concentrations (Kp) are required to estimate the brain penetration potential of a new drug entity. In this work, we developed machine learning models to predict in vivo compound brain penetration (as LogKp) from chemical structure. Our results show the benefit of including in vitro experimental data as auxiliary tasks in multi-task graph neural network (MT-GNN) models. MT-GNNs outperformed single-task (ST) models solely trained on in vivo brain penetration data. The best-performing MT-GNN regression model achieved a coefficient of determination of 0.42 and a mean absolute error of 0.39 (2.5-fold) on a prospective validation set and outperformed all tested ST models. To facilitate decision-making, compounds were classified into brain-penetrant or non-penetrant, achieving a Matthew's correlation coefficient of 0.66. Taken together, our findings indicate that the inclusion of in vitro assay data as MT-GNN auxiliary tasks improves in vivo brain penetration predictions and prospective compound prioritization.

Application of machine learning models for property prediction to targeted protein degraders.

Machine learning (ML) systems can model quantitative structure-property relationships (QSPR) using existing experimental data and make property predictions for new molecules. With the advent of modalities such as targeted protein degraders (TPD), the applicability of QSPR models is questioned and ML usage in TPD-centric projects remains limited. Herein, ML models are developed and evaluated for TPDs' property predictions, including passive permeability, metabolic clearance, cytochrome P450 inhibition, plasma protein binding, and lipophilicity. Interestingly, performance on TPDs is comparable to that of other modalities. Predictions for glues and heterobifunctionals often yield lower and higher errors, respectively. For permeability, CYP3A4 inhibition, and human and rat microsomal clearance, misclassification errors into high and low risk categories are lower than 4% for glues and 15% for heterobifunctionals. For all modalities, misclassification errors range from 0.8% to 8.1%. Investigated transfer learning strategies improve predictions for heterobifunctionals. This is the first comprehensive evaluation of ML for the prediction of absorption, distribution, metabolism, and excretion (ADME) and physicochemical properties of TPD molecules, including heterobifunctional and molecular glue sub-modalities. Taken together, our investigations show that ML-based QSPR models are applicable to TPDs and support ML usage for TPDs' design, to potentially accelerate drug discovery.

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.

Safety screening in early drug discovery: An optimized assay panel.

BACKGROUND: Several factors contribute to the development failure of novel pharmaceuticals, one of the most important being adverse effects in pre-clinical and clinical studies. Early identification of off-target compound activity can reduce safety-related attrition in development. In vitro profiling of drug candidates against a broad range of targets is an important part of the compound selection process. Many compounds are synthesized during early drug discovery, making it necessary to assess poly-pharmacology at a limited number of targets. This paper describes how a rational, statistical-ranking approach was used to generate a cost-effective, optimized panel of assays that allows selectivity focused structure-activity relationships to be explored for many molecules. This panel of 50 targets has been used to routinely screen Roche small molecules generated across a diverse range of therapeutic targets. Target hit rates from the Bioprint® database and internal Roche compounds are discussed. We further describe an example of how this panel was used within an anti-infective project to reduce in vivo testing. METHOD: To select the optimized panel of targets, IC50 values of compounds in the BioPrint® database were used to identify assay "hits" i.e. IC50 ≤ 1 µM in 123 different in vitro pharmacological assays. If groups of compounds hit the same targets, the target with the higher hit rate was selected, while others were considered redundant. Using a step-wise analysis, an assay panel was identified to maximize diversity and minimize redundancy. Over a five-year period, this panel of 50 off-targets was used to screen ≈1200 compounds synthesized for Roche drug discovery programs. Compounds were initially tested at 10 µM and hit rates generated are reported. Within one project, the number of hits was used to refine the choice of compounds being assessed in vivo. RESULTS: 95% of compounds from the BioPrint® panel were identified within the top 47-ranked assays. Based on this analytical approach and the addition of three targets with established safety concerns, a Roche panel was created for external screening. hERG is screened internally and not included in this analysis. Screening at 10 µM in the Roche panel identified that adenosine A3 and 5HT2B receptors had the highest hit rates (~30%), with 50% of the targets having a hit rate of ≤4%. An anti-infective program identified that a high number of hits in the Roche panel was associated with mortality in 19 mouse tolerability studies. To reduce the severity and number of such studies, future compound selections integrated the panel hit score into the selection process for in vivo studies. It was identified that compounds which hit less targets in the panel and had free plasma exposures of ~2 µM were generally better tolerated. DISCUSSION: This paper describes how an optimized panel of 50 assays was selected on the basis of hit similarity at 123 targets. This reduced panel, provides a cost-effective screening panel for assessing compound promiscuity, whilst also including many safety-relevant targets. Frequent use of the panel in early drug discovery has provided promiscuity and safety-relevant information to inform pre-clinical drug development at Roche.

Enhancement of drug absorption by noncharged detergents through membrane and P-glycoprotein binding.

Noncharged detergents are used as excipients in drug formulations. Until recently, they were considered as inert compounds, enhancing drug absorption essentially by improving drug solubility. However, many detergents insert into lipid membranes, although to different extents, and change the lateral packing density of membranes at high concentrations. Moreover, they bind to the efflux transporter P-glycoprotein (P-gp) and most likely to related transporters and metabolising enzymes with overlapping substrate specificities. If their affinity to P-gp is higher than that of the coadministered drug they act as modulators or inhibitors of P-gp and enhance drug absorption. Inhibition of P-gp and related proteins can, however, cause severe side effects. This paper first reviews the membrane binding propensity of different noncharged detergents (including poloxamers) and discusses their ability to bind to P-gp. Second, literature data on drug uptake enhancement by noncharged detergents, obtained in vivo and in vitro, are analysed at the molecular level. The present analysis provides the tools for an approximate and simple prior estimate of the membrane and P-gp binding ability of noncharged detergents based on a modular binding approach.

Can we discover pharmacological promiscuity early in the drug discovery process?

The term 'pharmacological promiscuity' describes the activity of a single compound against multiple targets. When undesired, promiscuity is a major safety concern that needs to be detected as early as possible in the drug discovery process. The analysis of large datasets reveals that the majority of promiscuous compounds are characterized by recognizable molecular properties and structural motifs, the most important one being a basic center with a pK(a)(B)>6. These compounds interact with a small set of targets such as aminergic GPCRs; some of these targets attract surprisingly high hit rates. In this review, we discuss current trends in the assessment of pharmacological promiscuity and propose strategies to enable early detection and mitigation.

Coexistence of passive and carrier-mediated processes in drug transport.

The permeability of biological membranes is one of the most important determinants of the pharmacokinetic processes of a drug. Although it is often accepted that many drug substances are transported across biological membranes by passive transcellular diffusion, a recent hypothesis speculated that carrier-mediated mechanisms might account for the majority of membrane drug transport processes in biological systems. Based on evidence of the physicochemical characteristics and of in vitro and in vivo findings for marketed drugs, as well as results from real-life discovery and development projects, we present the view that both passive transcellular processes and carrier-mediated processes coexist and contribute to drug transport activities across biological membranes.

In silico prediction of blood-brain barrier permeation using the calculated molecular cross-sectional area as main parameter.

The cross-sectional area, AD, of a compound oriented in an amphiphilic gradient such as the air-water or lipid-water interface has previously been shown to be crucial for membrane partitioning and permeation, respectively. Here, we developed an algorithm that determines the molecular axis of amphiphilicity and the cross-sectional area, ADcalc, perpendicular to this axis. Starting from the conformational ensemble of each molecule, the three-dimensional conformation selected as the membrane-binding conformation was the one with the smallest cross-sectional area, ADcalcM, and the strongest amphiphilicity. The calculated, ADcalcM, and the measured, AD, cross-sectional areas correlated linearly (n=55, slope, m=1.04, determination coefficient, r2=0.95). The calculated cross-sectional areas, ADcalcM, were then used together with the calculated octanol-water distribution coefficients, log D7.4, of the 55 compounds (with a known ability to permeate the blood-brain barrier) to establish a calibration diagram for the prediction of blood-brain barrier permeation. It yielded a limiting cross-sectional area (ADcalcM=70 A2) and an optimal range of octanol-water distribution coefficients (-1.4

Halogenation of drugs enhances membrane binding and permeation.

Halogenation of drugs is commonly used to enhance membrane binding and permeation. We quantify the effect of replacing a hydrogen residue by a chlorine or a trifluoromethyl residue in position C-2 of promazine, perazine, and perphenazine analogues. Moreover, we investigate the influence of the position (C-6 and C-7) of residue CF(3) in benzopyranols. The twelve drugs are characterized by surface activity measurements, which yield the cross-sectional area, the air-water partition coefficient, and the critical micelle concentration. By using the first two parameters (A(D) and K(aw)) and the appropriate membrane packing density, the lipid-water partition coefficients, are calculated in excellent agreement with the lipid-water partition coefficients measured by means of isothermal titration calorimetry for small unilamellar vesicles of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine. Replacement of a hydrogen residue by a chlorine and a trifluoromethyl residue enhances the free energy of partitioning into the lipid membrane, on average by deltaG(lw) approximately -1.3 or -4.5 kJ mol(-1), respectively, and the permeability coefficient by a factor of approximately 2 or approximately 9, respectively. Despite exhibiting practically identical hydrophobicities, the two benzopyranol analogues differ in their permeability coefficients by almost an order of magnitude; this is due to their different cross-sectional areas at the air-water and lipid-water interfaces.

Design and synthesis of novel imidazo[1,2-a]quinoxalines as PDE4 inhibitors.

New imidazo[1,2-a]quinoxaline derivatives have been synthesised by condensation of an appropriate alpha-aminoalcohol with a quinoxaline followed by intramolecular cyclisation and nucleophilic substitutions. Their phosphodiesterase inhibitory activities have been assessed on a preparation of the PDE4 isoform purified from a human alveolar epithelial cell line (A549). These studies showed potent inhibitory properties that emphasize the importance of a methyl amino group at position 4 and a weakly hindered group at position 1.