Reactivities of acrylamide warheads toward cysteine targets: a QM/ML approach to covalent inhibitor design.

Covalent inhibition offers many advantages over non-covalent inhibition, but covalent warhead reactivity must be carefully balanced to maintain potency while avoiding unwanted side effects. While warhead reactivities are commonly measured with assays, a computational model to predict warhead reactivities could be useful for several aspects of the covalent inhibitor design process. Studies have shown correlations between covalent warhead reactivities and quantum mechanic (QM) properties that describe important aspects of the covalent reaction mechanism. However, the models from these studies are often linear regression equations and can have limitations associated with their usage. Applications of machine learning (ML) models to predict covalent warhead reactivities with QM descriptors are not extensively seen in the literature. This study uses QM descriptors, calculated at different levels of theory, to train ML models to predict reactivities of covalent acrylamide warheads. The QM/ML models are compared with linear regression models built upon the same QM descriptors and with ML models trained on structure-based features like Morgan fingerprints and RDKit descriptors. Experiments show that the QM/ML models outperform the linear regression models and the structure-based ML models, and literature test sets demonstrate the power of the QM/ML models to predict reactivities of unseen acrylamide warhead scaffolds. Ultimately, these QM/ML models are effective, computationally feasible tools that can expedite the design of new covalent inhibitors.

Nonclassical Zwitterions as a Design Principle to Reduce Lipophilicity without Impacting Permeability.

The ionization of bioactive molecules impacts many ADME-relevant physicochemical properties, in particular, solubility, lipophilicity, and permeability. Ampholytes contain both acidic and basic groups and are distinguished as ordinary ampholytes and zwitterions. An influential review states that zwitterions only exist if the acidic pKa is significantly lower than the basic pKa. Through concordance of measured and calculated pKa and log P, we show that the zwitterionic behavior of several marketed drugs and natural products occurs despite a low or negative ΔpKa. These nonclassical zwitterions are characterized by a weak acidic and basic pKa and conjugation through an extended aromatic system, often including pseudorings via intramolecular hydrogen bonds. In contrast to most classical zwitterions, nonclassical zwitterions can exhibit excellent permeability. As permeability and lipophilicity are typically correlated, the combination of low lipophilicity and high permeability makes nonclassical zwitterions an attractive design principle in medicinal chemistry.

Quantitative and functional characterisation of extracellular vesicles after passive loading with hydrophobic or cholesterol-tagged small molecules.

Extracellular vesicles (EVs) are nanosized intercellular messengers that bear enormous application potential as biological drug delivery vehicles. Much progress has been made for loading or decorating EVs with proteins, peptides or RNAs using genetically engineered donor cells, but post-isolation loading with synthetic drugs and using EVs from natural sources remains challenging. In particular, quantitative and unambiguous data assessing whether and how small molecules associate with EVs versus other components in the samples are still lacking. Here we describe the systematic and quantitative characterisation of passive EV loading with small molecules based on hydrophobic interactions - either through direct adsorption of hydrophobic compounds, or by membrane anchoring of hydrophilic ligands via cholesterol tags. As revealed by single vesicle imaging, both ligand types bind to CD63 positive EVs (exosomes), however also non-specifically to other vesicles, particles, and serum proteins. The hydrophobic compounds Curcumin and Terbinafine aggregate on EVs with no apparent saturation up to 106-107 molecules per vesicle as quantified by liquid chromatography - high resolution mass spectrometry (LC-HRMS). For both compounds, high density EV loading resulted in the formation of a population of large, electron-dense vesicles as detected by quantitative cryo-transmission electron microscopy (TEM), a reduced EV cell uptake and a toxic gain of function for Curcumin-EVs. In contrast, cholesterol tagging of a hydrophilic mdm2-targeted cyclic peptide saturated at densities of ca 104-105 molecules per vesicle, with lipidomics showing addition to, rather than replacement of endogenous cholesterol. Cholesterol anchored ligands did not change the EVs' size or morphology, and such EVs retained their cell uptake activity without inducing cell toxicity. However, the cholesterol-anchored ligands were rapidly shed from the vesicles in presence of serum. Based on these data, we conclude that (1) both methods allow loading of EVs with small molecules but are prone to unspecific compound binding or redistribution to other components if present in the sample, (2) cholesterol anchoring needs substantial optimization of formulation stability for in vivo applications, whereas (3) careful titration of loading densities is warranted when relying on hydrophobic interactions of EVs with hydrophobic compounds to mitigate changes in physicochemical properties, loss of EV function and potential cell toxicity.

Lessons for Oral Bioavailability: How Conformationally Flexible Cyclic Peptides Enter and Cross Lipid Membranes.

Cyclic peptides extend the druggable target space due to their size, flexibility, and hydrogen-bonding capacity. However, these properties impact also their passive membrane permeability. As the "journey" through membranes cannot be monitored experimentally, little is known about the underlying process, which hinders rational design. Here, we use molecular simulations to uncover how cyclic peptides permeate a membrane. We show that side chains can act as "molecular anchors", establishing the first contact with the membrane and enabling insertion. Once inside, the peptides are positioned between headgroups and lipid tails─a unique polar/apolar interface. Only one of two distinct orientations at this interface allows for the formation of the permeable "closed" conformation. In the closed conformation, the peptide crosses to the lower leaflet via another "anchoring" and flipping mechanism. Our findings provide atomistic insights into the permeation process of flexible cyclic peptides and reveal design considerations for each step of the process.

Integrated and automated high-throughput purification of libraries on microscale.

We herein report the development of an automation platform for rapid purification and quantification of chemical libraries including reformatting of chemical matter to 10 mM DMSO stock solutions. This fully integrated workflow features tailored conditions for preparative reversed-phase (RP) HPLC-MS on microscale based on analytical data, online fraction QC and CAD-based quantification as well as automated reformatting to enable rapid purification of chemical libraries. This automated workflow is entirely solution-based, eliminating the need to weigh or handle solids. This increases process efficiency and creates a link between high-throughput synthesis and profiling of novel chemical matter with respect to biological and physicochemical properties in relevant assays.

Polar/apolar interfaces modulate the conformational behavior of cyclic peptides with impact on their passive membrane permeability.

Cyclic peptides have the potential to vastly extend the scope of druggable proteins and lead to new therapeutics for currently untreatable diseases. However, cyclic peptides often suffer from poor bioavailability. To uncover design principles for permeable cyclic peptides, a promising strategy is to analyze the conformational dynamics of the peptides using molecular dynamics (MD) and Markov state models (MSMs). Previous MD studies have focused on the conformational dynamics in pure aqueous or apolar environments to rationalize membrane permeability. However, during the key steps of the permeation through the membrane, cyclic peptides are exposed to interfaces between polar and apolar regions. Recent studies revealed that these interfaces constitute the free energy minima of the permeation process. Thus, a deeper understanding of the behavior of cyclic peptides at polar/apolar interfaces is desired. Here, we investigate the conformational and kinetic behavior of cyclic decapeptides at a water/chloroform interface using unbiased MD simulations and MSMs. The distinct environments at the interface alter the conformational equilibrium as well as the interconversion kinetics of cyclic peptide conformations. For peptides with low population of the permeable conformation in aqueous solution, the polar/apolar interface facilitates the interconversion to the closed conformation, which is required for membrane permeation. Comparison to unbiased MD simulations with a POPC bilayer reveals that not only the conformations but also the orientations are relevant in a membrane system. These findings allow us to propose a permeability model that includes both 'prefolding' and 'non-prefolding' cyclic peptides - an extension that can lead to new design considerations for permeable cyclic peptides.

Effect of Flexibility, Lipophilicity, and the Location of Polar Residues on the Passive Membrane Permeability of a Series of Cyclic Decapeptides.

Cyclic peptides have received increasing attention over the recent years as potential therapeutics for "undruggable" targets. One major obstacle is, however, their often relatively poor bioavailability. Here, we investigate the structure-permeability relationship of 24 cyclic decapeptides that share the same backbone N-methylation pattern but differ in their side chains. The peptides cover a large range of values for passive membrane permeability as well as lipophilicity and solubility. To rationalize the observed differences in permeability, we extracted for each peptide the population of the membrane-permeable conformation in water from extensive explicit-solvent molecular dynamics simulations and used this as a metric for conformational rigidity or "prefolding." The insights from the simulations together with lipophilicity measurements highlight the intricate interplay between polarity/lipophilicity and flexibility/rigidity and the possible compensating effects on permeability. The findings allow us to better understand the structure-permeability relationship of cyclic peptides and extract general guiding principles.

ReSCoSS: a flexible quantum chemistry workflow identifying relevant solution conformers of drug-like molecules.

Conformational equilibria are at the heart of drug design, yet their energetic description is often hampered by the insufficient accuracy of low-cost methods. Here we present a flexible and semi-automatic workflow based on quantum chemistry, ReSCoSS, designed to identify relevant conformers and predict their equilibria across different solvent environments in the Conductor-like Screening Model for Real Solvents (COSMO-RS) framework. We demonstrate the utility and accuracy of the workflow through conformational case studies on several drug-like molecules from literature where relevant conformations are known. We further show that including ReSCoSS conformers significantly improves COSMO-RS based predictions of physicochemical properties over single-conformation approaches. ReSCoSS has found broad adoption in the in-house drug discovery and development work streams and has contributed to establishing quantum-chemistry methods as a strategic pillar in ligand discovery.

Prediction of pKa Using Machine Learning Methods with Rooted Topological Torsion Fingerprints: Application to Aliphatic Amines.

The acid-base dissociation constant, pKa, is a key parameter to define the ionization state of a compound and directly affects its biopharmaceutical profile. In this study, we developed a novel approach for pKa prediction using rooted topological torsion fingerprints in combination with five machine learning (ML) methods: random forest, partial least squares, extreme gradient boosting, lasso regression, and support vector regression. With a large and diverse set of 14 499 experimental pKa values, pKa models were developed for aliphatic amines. The models demonstrated consistently good prediction statistics and were able to generate accurate prospective predictions as validated with an external test set of 726 pKa values (RMSE 0.45, MAE 0.33, and R2 0.84 by the top model). The factors that may affect prediction accuracy and model applicability were carefully assessed. The results demonstrated that rooted topological torsion fingerprints coupled with ML methods provide a promising approach for developing accurate pKa prediction models.

High accuracy quantum-chemistry-based calculation and blind prediction of macroscopic pKa values in the context of the SAMPL6 challenge.

Recent advances in the development of low-cost quantum chemical methods have made the prediction of conformational preferences and physicochemical properties of medium-sized drug-like molecules routinely feasible, with significant potential to advance drug discovery. In the context of the SAMPL6 challenge, macroscopic pKa values were blindly predicted for a set of 24 of such molecules. In this paper we present two similar quantum chemical based approaches based on the high accuracy calculation of standard reaction free energies and the subsequent determination of those pKa values via a linear free energy relationship. Both approaches use extensive conformational sampling and apply hybrid and double-hybrid density functional theory with continuum solvation to calculate free energies. The blindly calculated macroscopic pKa values were in excellent agreement with the experiment.

Developing Collaborative QSAR Models Without Sharing Structures.

It is widely understood that QSAR models greatly improve if more data are used. However, irrespective of model quality, once chemical structures diverge too far from the initial data set, the predictive performance of a model degrades quickly. To increase the applicability domain we need to increase the diversity of the training set. This can be achieved by combining data from diverse sources. Public data can be easily included; however, proprietary data may be more difficult to add due to intellectual property concerns. In this contribution, we will present a method for the collaborative development of linear regression models that addresses this problem. The method differs from other past approaches, because data are only shared in an aggregated form. This prohibits access to individual data points and therefore avoids the disclosure of confidential structural information. The final models are equivalent to models that were built with combined data sets.

Benefit of Retraining pKa Models Studied Using Internally Measured Data.

The ionization state of drugs influences many pharmaceutical properties such as their solubility, permeability, and biological activity. It is therefore important to understand the structure property relationship for the acid-base dissociation constant pKa during the lead optimization process to make better-informed design decisions. Computational approaches, such as implemented in MoKa, can help with this; however, they often predict with too large error especially for proprietary compounds. In this contribution, we look at how retraining helps to greatly improve prediction error. Using a longitudinal study with data measured over 15 years in a drug discovery environment, we assess the impact of model training on prediction accuracy and look at model degradation over time. Using the MoKa software, we will demonstrate that regular retraining is required to address changes in chemical space leading to model degradation over six to nine months.

What is modulating solubility in simulated intestinal fluids?

The aim of this study was to understand which parameters are responsible for the selective modulation of compounds solubility in simulated intestinal fluids. The solubility of 25 chemically diverse reference compounds was measured in simulated intestinal fluid (FaSSIF-V2) and in aqueous phosphate and maleate buffers. Electrostatic interactions between compounds and the bio-relevant medium components seem to explain the different solubility behavior observed for acids and bases. The solubility of ionized acids is not increased in FaSSIF-V2 probably due to electrostatic repulsions with the media components. Lipophilicity plays an important role but mainly for charged bases with a logP>4 (or logD(6.5)>1.9). When the aqueous solubility is mainly driven by lipophilicity, the FaSSIF-V2 components seem to improve the solubility of basic compounds to a greater extent than for compounds whose solubility is limited by crystal packing. These results suggest that ionization, lipophilicity and crystal packing play important but peculiar roles in controlling solubility in FaSSIF-V2 compared to that in aqueous buffer and this information could be useful to guide medicinal chemists and formulation scientists.