MELLODDY: Cross-pharma Federated Learning at Unprecedented Scale Unlocks Benefits in QSAR without Compromising Proprietary Information.

Federated multipartner machine learning has been touted as an appealing and efficient method to increase the effective training data volume and thereby the predictivity of models, particularly when the generation of training data is resource-intensive. In the landmark MELLODDY project, indeed, each of ten pharmaceutical companies realized aggregated improvements on its own classification or regression models through federated learning. To this end, they leveraged a novel implementation extending multitask learning across partners, on a platform audited for privacy and security. The experiments involved an unprecedented cross-pharma data set of 2.6+ billion confidential experimental activity data points, documenting 21+ million physical small molecules and 40+ thousand assays in on-target and secondary pharmacodynamics and pharmacokinetics. Appropriate complementary metrics were developed to evaluate the predictive performance in the federated setting. In addition to predictive performance increases in labeled space, the results point toward an extended applicability domain in federated learning. Increases in collective training data volume, including by means of auxiliary data resulting from single concentration high-throughput and imaging assays, continued to boost predictive performance, albeit with a saturating return. Markedly higher improvements were observed for the pharmacokinetics and safety panel assay-based task subsets.

Redox-neutral α-oxygenation of amines: reaction development and elucidation of the mechanism.

Cyclic secondary amines and 2-hydroxybenzaldehydes or related ketones react to furnish benzo[e][1,3]oxazine structures in generally good yields. This overall redox-neutral amine α-C-H functionalization features a combined reductive N-alkylation/oxidative α-functionalization and is catalyzed by acetic acid. In contrast to previous reports, no external oxidants or metal catalysts are required. Reactions performed under modified conditions lead to an apparent reductive amination and the formation of o-hydroxybenzylamines in a process that involves the oxidation of a second equivalent of amine. A detailed computational study employing density functional theory compares different mechanistic pathways and is used to explain the observed experimental findings. Furthermore, these results also reveal the origin of the catalytic efficiency of acetic acid in these transformations.

Metal-free α-amination of secondary amines: computational and experimental evidence for azaquinone methide and azomethine ylide intermediates.

We have performed a combined computational and experimental study to elucidate the mechanism of a metal-free α-amination of secondary amines. Calculations predicted azaquinone methides and azomethine ylides as the reactive intermediates and showed that iminium ions are unlikely to participate in these transformations. These results were confirmed by experimental deuterium-labeling studies and the successful trapping of the postulated azomethine ylide and azaquinone methide intermediates. In addition, computed barrier heights for the rate-limiting step correlate qualitatively with experimental findings.

Zwitterions and unobserved intermediates in organocatalytic Diels-Alder reactions of linear and cross-conjugated trienamines.

The Diels-Alder reactions of cyclic linear and cross-conjugated trienamines with oxindoles have been studied with density functional theory [M06-2X/def2-TZVPP/IEFPCM//B97D/6-31+G(d,p)/IEFPCM]. These reactions are found to proceed in a stepwise fashion. Computations revealed that these transformations involve complex mechanisms including zwitterionic intermediates and several unstable alternate cycloadducts arising from (2 + 2) cycloadditions and hetero-Diels-Alder reactions. The observed regio- and stereochemistry can be rationalized by a combination of kinetic and thermodynamic control.

Elucidation of Strong Cooperative Effects Caused by Dispersion Interactions in a Recognition-Mediated Diels-Alder Reaction.

Cooperative effects caused by dispersion interactions between dienes and dienophiles in a so-called self-replicating system have been evaluated by density functional theory. A variety of functionals is tested for the elucidation of reactant complex stabilities. Dispersion interactions between dienes and dienophiles result in additional stabilizations of up to 10 kcal mol(-1). These effects should be taken into account in future experimental studies and can be exploited to design more efficient systems.

Elucidating the origin of diastereoselectivity in a self-replicating system: selfishness versus altruism.

We have investigated a diastereoselective self-replicating system based on a cycloaddition of a fulvene derivative and a maleimide using a two-pronged approach of combining NMR spectroscopy with computational modelling. Two diastereomers are formed with identical rates in the absence of replication. When replication is enabled, one diastereomer takes over the resources as a "selfish" autocatalyst, while exploiting the competitor as a weak "altruist", resulting in a diastereoselectivity of 16:1. We applied 1D and 2D NMR spectroscopic techniques supported by ab initio chemical shifts as well as ab initio molecular dynamics simulations to study the structure and dynamics of the underlying network. This powerful combination allowed us to decipher the energetic and structural rationale behind the observed behaviour, while static computational methods currently used in the field did not.