(291d) A Novel Spectroscopy-Based Diagnostic Platform for Age-Related Diseases

The U.S. population is the oldest it has ever been, with the number of individuals aged 65 and older expected to surge from 58 million in 2022 to 82 million by 2050, marking a 47% increase. This unprecedented demographic shift is expected to coincide with a rise in patients with age-associated diseases (AADs), placing a growing burden on an already strained healthcare system. A major obstacle in treating many degenerative or age-associated diseases (AAD) is diagnosis before the onset of symptoms and irreversible progression. The foremost predictor in successfully treating any AAD or preventing related disability is an early diagnosis. However, several conditions like Alzheimer’s and atherosclerosis remain difficult to detect before the advent of serious symptoms. Biological aging manifests through the systemic accumulation of molecular- and cell-level damage. The permanent growth arrest and inflammatory senescence-associated secretory phenotype (SASP) of senescent cells is a unique mechanism or hallmark by which diverse age-related pathologies arise. The accumulation of senescent cells in several non-communicable AADs suggests a potential association, highlighting the possibility of these cells to serve as biomarkers for such conditions. Accurately detecting senescent cells in culture and in complex clinical samples remains challenging due to variations in surface markers and the unreliability of traditional assays across cell types. To solve this problem, we have developed a machine learning (ML)-assisted Raman spectroscopy platform for the detection and analysis of senescent cells. We generated senescent primary human dermal fibroblasts in vitro using 3 different induction methods. We validated the development of senescence using two traditional methods, RT-qPCR and SA-β-Gal staining. Senescent cell Raman spectra were collected across different induction methods using a WITec alpha confocal Raman microscope. We chemically defined senescence in pHDFs as a ratio of normalized intensity values corresponding to major biomolecules including carbohydrates, lipids, and proteins. Our platform reliably discerns between senescent and healthy cells. Our platform consists of a preprocessing pipeline to remove artifacts, baseline correct, normalize, and reduce the dimensionality of spectral data. We then evaluated a series of supervised (ANN & CNN) and unsupervised models (LDA, QDA, SVM) as methods of classifying various cell types. We achieved accuracies greater than 90% when attempting binary classification. Our preliminary results show Raman spectra can be used to accurately identify and cluster biologically distinct cell types, potentially providing specific information about disease type and severity. We avoid the pitfalls of previous senescence detection methods which require cell-type-specific reagents or have variable degrees of reliability depending on the tissue type being analyzed. Our approach holds significant promise as both a novel tool for cell analysis and as an early AAD-diagnostic. By understanding the distinct chemical compositions of senescent cell subpopulations, we can gain valuable insight into disease progression. With the deployment of this platform, researchers will, for the first time, be able to assess the efficacy of experimental anti-aging and AAD therapeutics with chemical detail in a single-cell, non-destructive, and live-cell compatible manner.