Abstract Drug discovery and testing are both time-consuming and costly. Leveraging single-cell transcriptomic data, we developed a machine learning framework to model patient cells using scRNA-seq data, aiming to infer gene regulatory networks (GRNs) that drive cell states and plasticity. Our computational algorithm predicts cellular responses to drug perturbations at the transcriptional level, even without experimental testing. We utilized this algorithm to identify drug candidates that could shift anti-inflammatory macrophages to a pro-inflammatory state for remodeling the tumor microenvironment, reactivate exhausted CD8 T cells, and convert immunosuppressive CD4 Treg cells into central memory CD4 cells to enhance anti-tumor activity. The predicted outcomes were validated experimentally through scRNA-seq, flow cytometry, and cytokine profiling. In human hepatocellular carcinoma (HCC), tumor infiltrating lymphocytes (TILs) treated with the predicted drugs showed a decreased FoxP3+ CD4 Treg population compared to untreated control TILs from the same HCC (n = 3). Separately, TILs treated with another set of predicted drugs showed higher levels of IFNG+ CD8 T cells than untreated controls (n = 3). Lastly, in human THP1 macrophages differentiated to an anti-inflammatory state, treatment resulted in increased levels of TNFa and IL1b secretion compared to untreated controls, indicating a phenotypic shift towards a pro-inflammatory state. We then tested the same drug on tumor associated macrophages (TAMs) from HCC and showed an increased CD64+ HLA-DR+ macrophage subset upon treatment. scRNA-seq analysis further revealed key pathways induced by the drugs, that resulted in the desired change in immune phenotype, offering insight into immune cell function and plasticity. In conclusion, this study introduced a machine-learning driven approach to identify candidate small molecules from scRNA-seq data that are capable of modulating macrophage and T cell activation, providing new avenues for targeted interventions in cancer. Citation Format: N. Jannah M. Nasir, John F. Ouyang, Valerie Chew, Owen Rackham. Applying Machine Learning to Identify Enhancers of Immune Cell Functionality for Targeted Cancer Therapies abstract. In: Proceedings of Frontiers in Cancer Science 2024; 2024 Nov 13-15; Singapore. Philadelphia (PA): AACR; Cancer Res 2025;85 (15Suppl): Abstract nr P54.
Nasir et al. (Fri,) studied this question.