Decoding Early Signs of Alzheimer's Disease in The Lateral Entorhinal Cortex Using Machine Learning

侧内嗅皮层（LEC）是阿尔茨海默氏症中最先受到影响的大脑区域之一

The lateral entorhinal cortex (LEC) is one of the first regions in the brain to be affected in Alzheimer’s disease, and is important for object recognition, odor discrimination and episodic memory. Hence, early AD symptoms such as misplacing objects, forgetting events and loss of smell could be due to LEC dysfunction. In order to understand how Aβ and tau accumulation impacts the LEC neurons, we will use two mouse models of AD: APP knockin (APP-KI) mice- expressing physiological levels of APP and EC- APP/Tau mice- expressing elevated levels of APP and tau in the EC. Both mouse models show selectively vulnerability in the LEC, making them ideal candidates to probe LEC function. Our preliminary data shows behavioral impairment in the EC-APP/Tau mice at 24 months and data on APP-KI show impairment at 18 months. In the proposal we will evaluate LEC function in the younger mice in order to detect neuronal changes prior to behavioral deficits. We will record LEC activity with silicon probes and test responses towards objects, odors and passage of time. Using computational approach such as machine learning, we will determine if ensemble properties of LEC neurons are affected by tau and Aβ. We hypothesize that APP in the LEC of APP-KI mice will make the neurons dysfunctional which will be evident with poor decoding accuracy for objects, odors and temporal epochs. In the EC-APP/Tau mice, combined effect of Aβ and tau will make the dysfunction worse and affect the decoding accuracy further allowing better prediction of early symptoms of Alzheimer’s disease. The proposal brings together diverse fields (electrophysiology, pathology and computational neuroscience) applying large-scale recording techniques to record ensemble populations of neurons and develop analytical and predictive computational tests to interrogate function in a vulnerable brain region that is dysfunctional in Alzheimer’s disease.