Creating Chemiluminescence Signature Arrays Coupled with Machine Learning for Alzheimer's Disease Serum Diagnosis.

Abstract

Although omics and multi-omics approaches are the most used methods to create signature arrays for liquid biopsy, the high cost of omics technologies still largely limits their wide applications for point-of-care. Inspired by the bat echolocation mechanism, we propose an "echoes" approach for creating chemiluminescence signatures via screening of a compound library, and serum samples of Alzheimer's disease (AD) were used for our proof-of-concept study. We first demonstrated the discrepancy in physicochemical properties between AD and healthy control serums. On this basis, we developed a simple, cost-effective, and versatile platform termed UNICODE (UNiversal Interaction of Chemiluminescence echOes for Disease Evaluation). The UNICODE platform consists of a "bat" probe, which generates different chemiluminescence intensities upon interacting with various substrates, and a panel/array of "flag" molecules that are selected from library screening. The UNICODE array could enable the reflecting/"echoing" of the signatures of various serum components and intact physicochemical interactions between serum substrates. In this study, we screened a library of over 1,000 small molecules and identified 12 "flag" molecules (top 12) that optimally depict the differences between AD and healthy control serums. Finally, we employed the top 12 array to conduct tests on serum samples and utilized machine learning methods to optimize detection performance. We successfully distinguished AD serums, achieving the highest area under the curve of 90.24% with the random forest method. Our strategy could provide new insights into biofluid abnormality and prototype tools for developing liquid biopsy diagnoses for AD and other diseases.