Small cell lung cancer (SCLC) is an aggressive, metastatic malignancy. Despite treatment regimens comprising combinations of chemo and immunotherapies, patient outcomes remain poor, with a 5 year survival of 1-10%. Interactions between cancer and immune cells drive treatment resistance in many types of cancers, such as ovarian and brain. It is likely that such a pathophysiological mechanism exists in SCLC, and underlies poor response of this cancer to treatment. Recent studies have identified high levels of immune cell infiltrate in SCLC. However, it remains unknown how these interactions change with treatment and if chemotherapy may unmask cellular interactions or signaling pathways that can be targeted for therapeutic benefit. Here, we aim to further analyze a public SCLC single-cell transcriptomic dataset containing treated and untreated tumours to generate a graph representation, followed by predictions of tumour-immune interactions across both conditions via graph neural network (GNN) approaches. Using cell-type annotations present in the published data, we will identify the cellular basis of tumour-immune interactions in both treatment conditions. Next, we will identify gene expression changes and cell interactions within SCLC cells via different forms of GNNs and machine learning algorithms for prediction and biomarker identification. Subsequently, we will apply pathway analysis to test how activation of molecular signalling pathways changes with treatment. Finally, we will conduct a literature search for small molecules known to disrupt these cellular interactions and molecular pathways and propose lead compounds that can be further tested by the lung cancer research community in pre-clinical experimental models.

We will conclude our study by performing a semi-automatic literature search for small molecules known to disrupt the cellular interactions and molecular signalling pathways we identify as potential therapeutic strategies for SCLC patients. Our goal is to generate a set of lead compounds that can be further explored by the cancer research community in preclinical models, and demonstrate an approach to rational treatment development that may be applied to other disease types in the future.