Combining radiomics and mathematical modeling to elucidate mechanisms of resistance to immune checkpoint blockade in non-small cell lung cancer

Abstract

Immune therapies have shown promise in a number of cancers, and clinical trials using the anti-PD-L1/PD-1 checkpoint inhibitor in lung cancer have been successful for a number of patients. However, some patients either do not respond to the treatment or have cancer recurrence after an initial response. It is not clear which patients might fall into these categories or what mechanisms are responsible for treatment failure. To explore the different underlying biological mechanisms of resistance, we created a spatially explicit mathematical model with a modular framework. This construction enables different potential mechanisms to be turned on and off in order to adjust specific tumor and tissue interactions to match a specific patient's disease. In parallel, we developed a software suite to identify significant computed tomography (CT) imaging features correlated with outcome using data from an anti-PDL-1 checkpoint inhibitor clinical trial for lung cancer and a tool that extracts these features from both patient CT images and “virtual CT” images created from the cellular density profile of the model. The combination of our two toolkits provides a framework that feeds patient data through an iterative pipeline to identify predictive imaging features associated with outcome, whilst at the same time proposing hypotheses about the underlying resistance mechanisms.

Changes in CT Radiomic Features Associated with Lymphocyte Distribution Predict Overall Survival and Response to Immunotherapy in Non–Small Cell Lung Cancer

“…Previous studies have also used radiomics to predict response to ICIs (15,24,62,63). Response to immunotherapy was negatively correlated with tumor convexity and positively correlated with edgeto-core size ratio on CT scans of patients with NSCLC (64). In a study by Tang and colleagues (63), favorable outcome group of immune response characterized by low CT intensity and high heterogeneity exhibited low PDL1 and high CD3 infiltration, suggestive of a favorable immune activated state.…”

CLR01 protects dopaminergic neurons in vitro and in mouse models of Parkinson’s disease

“…Mathematical in silico models have been previously used to predict and model pharmacological and biological processes [21][22][23][24] . We first administered radiolabeled CLR01 systemically to 12month-old animals to measure the blood-brain penetration of the compound to inform an in silico model that would allow us to determine whether a monthly dose of 40 µg/kg/day, which was previously used 8 , would provide favorable pharmacokinetic profile in the brain.…”

From cells to tissue: How cell scale heterogeneity impacts glioblastoma growth and treatment response

“…At the imaging scale, spatial variations can be quantified to reveal habitats and predict treatment response. Radiomic imaging does just that, because nuances in the shape, morphology, and texture of tumor density maps gives more information than size dynamics alone [3,[6][7][8]18].…”

“…However, it is often the case that patients with similar growth patterns determined with MRI will have different post-treatment kinetics. While patient data at smaller scales, such as histological and genetic profiling, is known to be generally prognostic, its connection to optimal therapeutics and clinical imaging remains an active area of research [3][4][5][6][7][8]. In this work, we investigate how phenotypic heterogeneity at the cell scale affects tumor growth and treatment response at the imaging scale by quantitatively matching multiscale data from an experimental rat model of GBM to a mechanistic computational model.…”