Hoshino Ayuko, PhD, MS; Lav Varshney, PhD; Alexandru Hanganu, PhD
Competition Sponsor: Japan Agency for Medical Research and Development
Exosomes are small packages of proteins and genetic material that are released into the blood from various cell types and organ systems. They then flow through the blood and into other cells and organs, as a kind of cell-to-cell communication system. It has recently been discovered, however, that the messages carried within exosomes may cause numerous age-related diseases including cancer and Alzheimerʼs disease (AD). In this work, we aim to characterize the trajectory of how the nature of messages within exosomes change both in healthy aging and in patients that have AD. To do so, we will first use statistical characterizations of age-related changes for the healthy and AD trajectories, and try to interpret the biological significance of specific proteins. Our preliminary data already has shown that the total exosomal protein concentration increases with onset of AD. As a second approach we aim to understand what is happening at a deeper biological level. Taking inspiration from artificial intelligence (AI) approaches in modern natural language processing that use deep learning. The idea is to train a protein language model using large corpora of amino acid sequences to understand the grammar of proteins, and then fine-tune it to classify the cell types and organ systems from where exosomes are coming from. This will allow us to understand the language of the messages in exosomes, as well as the communication paths that exosomes take as they flow through the body. With this understanding, we will be able to explain the biological pathways by which exosomes act, how these pathways evolve with healthy aging, and further to make early predictions for diseases such as Alzheimerʼs disease.