PROJECT 2: Development and Refinement of Predictive Models for Designing Immunotherapy Combination Treatments

Project 2 Project Summary Cutaneous melanoma became an early example for treatment with targeted therapy with the clinical development of the first BRAFV600MUT-specific inhibitor (BRAFi), vemurafenib1. Resistance to BRAFi is common2, and was initially ascribed to cancer cell intrinsic factors that reactivate MAPK pathway signaling3–7. BRAFi in combination with MEKi8 was developed to combat such resistance, but only a quarter of patients treated with this combination survive for five years9. In fact, recent data suggest that cancer cell-extrinsic factors, including immune factors3,10–13, can play important roles in resistance development to MAPK pathway inhibitors, thus highlighting the role of the tumor-immune microenvironment (TIME). Strikingly, in syngeneic melanoma models that develop resistance against both MAPKi and immune checkpoint blockade (ICB), lead-in ICB can ‘prime’ both the primary tumor and distal metastases for eradication when the ICB is subsequently combined with MAPKi14. While this suggests that immune based strategies, such as ICB or adoptive cell therapy (ACT), can serve as sequential combinatorial agents to prevent MAPKi resistance. However, it also significantly complicates the design of candidate treatment regimens, since multiple sequences and sequence timings need to be tested. This can make clinical trials design impractical. We propose to develop methods that apply iterative and active learning to deep phenotyping with spatial and temporal multi-omics assays to yield predictive in silico models that can provide guidance for designing sequential immunotherapy - targeted inhibitor combination therapies. . A key element of Project 2 is the iterative development of multiscale Agent Based Models (ABMs) as a virtual representation of the TIME. ABMs are initially constructed from existing data, including preliminary results from biobanked tumor specimens and public omics data bases, and from our extensive experience within the Cancer Genome Atlas (TCGA). They are then evolved through a systems biology-inspired iterative cycle of quantitative experimentation, analysis, modeling, and validation, drawing from experimental data from both Projects 1 and 2.