Develop SASP-AI: a generative model for predicting the paracrine effects of the SASP

National Institute On Aging

Investigators

Abstract

The senescence-associated secretory phenotype (SASP) contributes to aging-related diseases and displays high heterogeneity across inducers and cell types. Despite its known pleiotropic effects on neighboring tissue, the link between SASP composition and paracrine response remains unclear. Stress-induced remodeling further complicates this landscape, emphasizing the need to define how specific SASP profiles drive distinct cellular outcomes. To address this, we screen SASP factors in parallel and assess their effects by multiplexed single-cell RNA sequencing. We expose cells to individual factors across doses and capture transcriptomic responses. The resulting data reveal transcriptomic signatures associated with specific SASP components. A neural network is then trained to predict cell responses from basal transcriptomic cell state and SASP input profiles. We evaluate the model accuracy using matched transcriptomic data for held out SASP input profiles and targeted low throughput transcriptomic experiments.