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.

Development of in silico models to predict viscosity and mouse clearance using a comprehensive analytical data set collected on 83 scaffold-consistent monoclonal antibodies.

Biologic drug discovery pipelines are designed to deliver protein therapeutics that have exquisite functional potency and selectivity while also manifesting biophysical characteristics suitable for manufacturing, storage, and convenient administration to patients. The ability to use computational methods to predict biophysical properties from protein sequence, potentially in combination with high throughput assays, could decrease timelines and increase the success rates for therapeutic developability engineering by eliminating lengthy and expensive cycles of recombinant protein production and testing. To support development of high-quality predictive models for antibody developability, we designed a sequence-diverse panel of 83 effector functionless IgG1 antibodies displaying a range of biophysical properties, produced and formulated each protein under standard platform conditions, and collected a comprehensive package of analytical data, including in vitro assays and in vivo mouse pharmacokinetics. We used this robust training data set to build machine learning classifier models that can predict complex protein behavior from these data and features derived from predicted and/or experimental structures. Our models predict with 87% accuracy whether viscosity at 150 mg/mL is above or below a threshold of 15 centipoise (cP) and with 75% accuracy whether the area under the plasma drug concentration-time curve (AUC0-672 h) in normal mouse is above or below a threshold of 3.9 × 106 h x ng/mL.

Chemi-Net: A Molecular Graph Convolutional Network for Accurate Drug Property Prediction.

Absorption, distribution, metabolism, and excretion (ADME) studies are critical for drug discovery. Conventionally, these tasks, together with other chemical property predictions, rely on domain-specific feature descriptors, or fingerprints. Following the recent success of neural networks, we developed Chemi-Net, a completely data-driven, domain knowledge-free, deep learning method for ADME property prediction. To compare the relative performance of Chemi-Net with Cubist, one of the popular machine learning programs used by Amgen, a large-scale ADME property prediction study was performed on-site at Amgen. For all 13 data sets, Chemi-Net resulted in higher R2 values compared with the Cubist benchmark. The median R2 increase rate over Cubist was 26.7%. We expect that the significantly increased accuracy of ADME prediction seen with Chemi-Net over Cubist will greatly accelerate drug discovery.

Fragment based drug discovery: practical implementation based on ¹9F NMR spectroscopy.

Fragment based drug discovery (FBDD) is a widely used tool for discovering novel therapeutics. NMR is a powerful means for implementing FBDD, and several approaches have been proposed utilizing (1)H-(15)N heteronuclear single quantum coherence (HSQC) as well as one-dimensional (1)H and (19)F NMR to screen compound mixtures against a target of interest. While proton-based NMR methods of fragment screening (FBS) have been well documented and are widely used, the use of (19)F detection in FBS has been only recently introduced (Vulpetti et al. J. Am. Chem. Soc.2009, 131 (36), 12949-12959) with the aim of targeting "fluorophilic" sites in proteins. Here, we demonstrate a more general use of (19)F NMR-based fragment screening in several areas: as a key tool for rapid and sensitive detection of fragment hits, as a method for the rapid development of structure-activity relationship (SAR) on the hit-to-lead path using in-house libraries and/or commercially available compounds, and as a quick and efficient means of assessing target druggability.

Discovery and initial SAR of 3-(1H-benzo[d]imidazol-2-yl)pyridin-2(1H)-ones as inhibitors of insulin-like growth factor 1-receptor (IGF-1R).

The discovery and synthesis of 3-(1H-benzo[d]imidazol-2-yl)pyridin-2(1H)-one inhibitors of insulin-like growth factor 1-receptor (IGF-1R) are presented. Installing amine containing side chains at the 4-position of pyridone ring significantly improved the enzyme potency. SAR and biological activity of these compounds is presented.

Novel 1H-(benzimidazol-2-yl)-1H-pyridin-2-one inhibitors of insulin-like growth factor I (IGF-1R) kinase.

A novel class of 1H-(benzimidazol-2-yl)-1H-pyridin-2-one inhibitors of insulin-like growth factor I (IGF-1R) kinase is described. This report discusses the SAR of 4-(2-hydroxy-2-phenylethylamino)-substituted pyridones with improved IGF-1R potency.

Discovery of a (1H-benzoimidazol-2-yl)-1H-pyridin-2-one (BMS-536924) inhibitor of insulin-like growth factor I receptor kinase with in vivo antitumor activity.

Compound 3 (BMS-536924), a novel small-molecule inhibitor of the insulin-like growth factor receptor kinase with equal potency against the insulin receptor is described. The in vitro and in vivo biological activity of this interesting compound is also reported.

Molecular determinants of vanilloid sensitivity in TRPV1.

Vanilloid receptor 1 (TRPV1), a membrane-associated cation channel, is activated by the pungent vanilloid from chili peppers, capsaicin, and the ultra potent vanilloid from Euphorbia resinifera, resiniferatoxin (RTX), as well as by physical stimuli (heat and protons) and proposed endogenous ligands (anandamide, N-arachidonyldopamine, N-oleoyldopamine, and products of lipoxygenase). Only limited information is available in TRPV1 on the residues that contribute to vanilloid activation. Interestingly, rabbits have been suggested to be insensitive to capsaicin and have been shown to lack detectable [(3)H]RTX binding in membranes prepared from their dorsal root ganglia. We have cloned rabbit TRPV1 (oTRPV1) and report that it exhibits high homology to rat and human TRPV1. Like its mammalian orthologs, oTRPV1 is selectively expressed in sensory neurons and is sensitive to protons and heat activation but is 100-fold less sensitive to vanilloid activation than either rat or human. Here we identify key residues (Met(547) and Thr(550)) in transmembrane regions 3 and 4 (TM3/4) of rat and human TRPV1 that confer vanilloid sensitivity, [(3)H]RTX binding and competitive antagonist binding to rabbit TRPV1. We also show that these residues differentially affect ligand recognition as well as the assays of functional response versus ligand binding. Furthermore, these residues account for the reported pharmacological differences of RTX, PPAHV (phorbol 12-phenyl-acetate 13-acetate 20-homovanillate) and capsazepine between human and rat TRPV1. Based on our data we propose a model of the TM3/4 region of TRPV1 bound to capsaicin or RTX that may aid in the development of potent TRPV1 antagonists with utility in the treatment of sensory disorders.