Building a Neural Network: Predicting Mice Behaviors from rCPT Activity

Overview

Machine learning and artificial intelligence have truly become some of the most groundbreaking technologies of modern time. Particularly, in the field of behavioral neuroscience, deep learning can be applied to the quantification of animal behaviors, which can unveil brain function in major physiological conditions. In this study, an advanced training platform, SLEAP (Social LEAP Estimates Animal Pose), was utilized to develop a neural network capable of tracking animal body part positions and predicting th e behavioral mechanisms of rodents performing a continuous performance test (rCPT). The rCPT paradigm requires experimental mice to accurately discriminate and respond to brief presentations of nontarget and target stimuli in order to earn a condensed milk reward. This study analyzed data from the rCPT activity of wild-type mice and mice lacking forebrain acetylcholine expression (that have deficits in attentional function). The model was developed using a bottom -up approach, in which select mice body parts were first labeled from overhead videos in order to train an artificial neural network. Thereafter, the modified network was able to accurately associate the marker labels with their corresponding animals. Learning classification algorithms were then run to quantify the specific mice behavior, which supported the association between mice position and attention deficit in the rCPT. Thus, the neural network developed in this experiment can be applied to mice models of diverse neurological impairments. Ultima tely, it can be further refined to test novel drug treatments in preclinical trials based on models of those behavioral conditions. Computational Drug Discovery for α-synuclein Inhibition in Parkinson’s Disease Neurodegeneration Jeremy Nashid Dr. Ronald E. McNair Academic High School, Jersey City, NJ Parkinson’s disease (PD) is a neurodegenerative disorder that leads to neuronal damage mainly because of the accumulation of misfolded alpha-synuclein proteins. This study uses a ligand docking approach using Schrödinger Maestro to identify potential compo unds with the highest potential to inhibit alpha -synuclein aggregation. A diverse set of small molecules and peptides, found in literature to be known to have some effect on PD, were docked to predict their interactions with alpha-synuclein. Before docking trials began, it was hypothesized that Rifampicin would be a powerful candidate, which was supported by its molecular properties. Next, site analysis pinpointed the most promising binding site on alpha -synuclein, enhancing the accuracy of subsequent trials. Finally, the molecular docking simulations and compounds like Lacmoid, Rosmarinic Acid, and Synuclean-Dexhibited the strongest docking scores, while molecules like Rifampicin, which had been hypothesized to be most effective, displayed no results. The data generated from this molecular docking study opens up potential therapeutic treatments for PD, and can guide future experiments in vitro to develop a cure for the neuronal damage related symptoms of PD.

Competition history

• JSHS 2024 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.