PathwayMap: Molecular Pathway Association with Self-Normalizing Neural Networks.

Drug discovery suffers from high attrition because compounds initially deemed as promising can later show ineffectiveness or toxicity resulting from a poor understanding of their activity profile. In this work, we describe a deep self-normalizing neural network model for the prediction of molecular pathway association and evaluate its performance, showing an AUC ranging from 0.69 to 0.91 on a set of compounds extracted from ChEMBL and from 0.81 to 0.83 on an external data set provided by Novartis. We finally discuss the applicability of the proposed model in the domain of lead discovery. A usable application is available via PlayMolecule.org .

Sleep in honey bees is affected by the herbicide glyphosate.

Sleep plays an essential role in both neural and energetic homeostasis of animals. Honey bees (Apis mellifera) manifest the sleep state as a reduction in muscle tone and antennal movements, which is susceptible to physical or chemical disturbances. This social insect is one of the most important pollinators in agricultural ecosystems, being exposed to a great variety of agrochemicals, which might affect its sleep behaviour. The intake of glyphosate (GLY), the herbicide most widely used worldwide, impairs learning, gustatory responsiveness and navigation in honey bees. In general, these cognitive abilities are linked with the amount and quality of sleep. Furthermore, it has been reported that animals exposed to sleep disturbances show impairments in both metabolism and memory consolidation. Consequently, we assessed the sleep pattern of bees fed with a sugar solution containing GLY (0, 25, 50 and 100 ng) by quantifying their antennal activity during the scotophase. We found that the ingestion of 50 ng of GLY decreased both antennal activity and sleep bout frequency. This sleep deepening after GLY intake could be explained as a consequence of the regenerative function of sleep and the metabolic stress induced by the herbicide.

An Artificial Intelligence Approach to Proactively Inspire Drug Discovery with Recommendations.

Artificial intelligence (AI) is becoming established in drug discovery. For example, many in the industry are applying machine learning approaches to target discovery or to optimize compound synthesis. While our organization is certainly applying these sorts of approaches, we propose an additional approach: using AI to augment human intelligence. We have been working on a series of recommendation systems that take advantage of our existing laboratory processes, both wet and computational, in order to provide inspiration to our chemists, suggest next steps in their work, and automate existing workflows. We will describe five such systems in various stages of deployment within the Novartis Institutes for BioMedical Research. While each of these systems addresses different stages of the discovery pipeline, all of them share three common features: a trigger that initiates the recommendation, an analysis that leverages our existing systems with AI, and the delivery of a recommendation. The goal of all of these systems is to inspire and accelerate the drug discovery process.

Learning and Its Neural Correlates in a Virtual Environment for Honeybees.

The search for neural correlates of operant and observational learning requires a combination of two (experimental) conditions that are very difficult to combine: stable recording from high order neurons and free movement of the animal in a rather natural environment. We developed a virtual environment (VE) that simulates a simplified 3D world for honeybees walking stationary on an air-supported spherical treadmill. We show that honeybees perceive the stimuli in the VE as meaningful by transferring learned information from free flight to the virtual world. In search for neural correlates of learning in the VE, mushroom body extrinsic neurons were recorded over days during learning. We found changes in the neural activity specific to the rewarded and unrewarded visual stimuli. Our results suggest an involvement of the mushroom body extrinsic neurons in operant learning in the honeybee (Apis mellifera).

Context odor presentation during sleep enhances memory in honeybees.

Sleep plays an important role in stabilizing new memory traces after learning [1-3]. Here we investigate whether sleep's role in memory processing is similar in evolutionarily distant species and demonstrate that a context trigger during deep-sleep phases improves memory in invertebrates, as it does in humans. We show that in honeybees (Apis mellifera), exposure to an odor during deep sleep that has been present during learning improves memory performance the following day. Presentation of the context odor during wake phases or novel odors during sleep does not enhance memory. In humans, memory consolidation can be triggered by presentation of a context odor during slow-wave sleep that had been present during learning [3-5]. Our results reveal that deep-sleep phases in honeybees have the potential to prompt memory consolidation, just as they do in humans. This study provides strong evidence for a conserved role of sleep-and how it affects memory processes-from insects to mammals.