RESUMEN
Alcohol is a widely used and abused substance. A major unresolved issue in the alcohol research field is determining which of the many alcohol target proteins identified to date is responsible for shaping each specific alcohol-related behavior. The large-conductance, calcium- and voltage-activated potassium channel (BK channel) is a conserved target of ethanol. Genetic manipulation of the highly conserved BKα channel influences alcohol-related behaviors across phylogenetically diverse species that include worm, fly, mouse, and man. A pharmacological tool that prevents alcohol's action at a single target, like the BK channel, would complement genetic approaches in the quest to define the behavioral consequences of alcohol at each target. To identify agents that specifically modulate the action of ethanol at the BK channel, we executed a high-throughput phagemid-display screen in combination with a Caenorhabditis elegans behavioral genetics assay. This screen selected a novel nonapeptide, LS10, which moderated acute ethanol intoxication in a BK channel-humanized C. elegans strain without altering basal behavior. LS10's action in vivo was dependent upon BK channel functional activity. Single-channel electrophysiological recordings in vitro showed that preincubation with a submicromolar concentration of LS10 restricted ethanol-induced changes in human BKα channel gating. In contrast, no substantial changes in basal human BKα channel function were observed after LS10 application. The results obtained with the LS10 peptide provide proof-of-concept evidence that a combined phagemid-display/behavioral genetics screening approach can provide novel tools for understanding the action of alcohol at the BK channel and how this, in turn, exerts influence over central nervous system function.
Asunto(s)
Etanol/farmacología , Canales de Potasio de Gran Conductancia Activados por el Calcio/metabolismo , Péptidos/metabolismo , Alcoholismo/metabolismo , Animales , Caenorhabditis elegans , Línea Celular , Células HEK293 , Humanos , Neuronas/efectos de los fármacos , Neuronas/metabolismo , XenopusRESUMEN
The rise of machine learning (ML) has created an explosion in the potential strategies for using data to make scientific predictions. For physical scientists wishing to apply ML strategies to a particular domain, it can be difficult to assess in advance what strategy to adopt within a vast space of possibilities. Here we outline the results of an online community-powered effort to swarm search the space of ML strategies and develop algorithms for predicting atomic-pairwise nuclear magnetic resonance (NMR) properties in molecules. Using an open-source dataset, we worked with Kaggle to design and host a 3-month competition which received 47,800 ML model predictions from 2,700 teams in 84 countries. Within 3 weeks, the Kaggle community produced models with comparable accuracy to our best previously published 'in-house' efforts. A meta-ensemble model constructed as a linear combination of the top predictions has a prediction accuracy which exceeds that of any individual model, 7-19x better than our previous state-of-the-art. The results highlight the potential of transformer architectures for predicting quantum mechanical (QM) molecular properties.