Project Summary Preclinical models of lung cancer are essential tools for researchers to understand cancer biology and develop therapeutic strategies. Choosing the most cost-effective preclinical model to answer a specific scientific question requires careful study of the existing molecular and pathological characterization of different models and human tumors, as molecular profiling reveals the orchestration of biological processes and pathological characterization informs the spatial composition of the tumor microenvironment. While preclinical models of lung cancer have been extensively characterized, the molecular data remain scattered, the pathology data have rarely been deposited, and no tool exists to evaluate the molecular and pathological agreement between preclinical models and human tumors. We have previously built a lung cancer explorer, which provides user-friendly integrative analytical tools to explore gene expression and clinical data from over 6,700 patients in 56 published datasets. Leveraging patient lung tumor pathology image archives, we developed algorithms and pipelines to perform histopathology digital staining and feature extraction from H&E images and identified interesting pathology features that predict outcome and response to targeted therapy. Extending these efforts to preclinical models, this proposal aims to develop an informatics platform integrating molecular and pathology data from various lung cancer preclinical models and patient tumors to assess preclinical model fidelity through comparative analyses. Specific Aim 1 will harmonize molecular profiling datasets from various lung cancer preclinical models. Statistical methods for cross-study validation and quality control will be implemented to ensure computational compatibility and to select appropriate datasets for analysis. Model-specific web applications will be built to support data exploration and analysis. Specific Aim 2 will perform histopathological and spatial transcriptomic characterization of tumors from in vivo models. We will network with lung cancer preclinical model investigators to solicit contributions of pathology images and samples for establishing a public image archive and for spatial molecular profiling experiments. Effective algorithms and pipelines for preclinical model pathology image analyses will be established. Specific Aim 3 will integrate data collected and harmonized in Aims 1 and 2 to construct an informatics platform for cross-model comparison and alignment to human tumors. This platform will allow users to review and download our processed molecular and pathology datasets and compare molecular and pathology profiles of preclinical models and patient tumors from multiple facets. We will share these resources with the lung cancer research community and solicit feedback to improve our platform. The successful implementation of this project will assemble the scattered molecular datasets, establish a large- scale public pathology image and spatial molecular profiling resource, and establish a user-friendly integrative fidelity assessment platform for lung cancer preclinical models.

Project Narrative: Preclinical models of lung cancer are essential tools for researchers to understand cancer biology and develop therapeutic strategies, but how they compare with each other and align with human tumors remains elusive. This study proposes to collect and harmonize molecular and pathology data from a wide range of lung cancer preclinical models. These data will be used to generate a user-friendly web application with a myriad of informatics tools for the broad lung cancer community to explore and assess whether preclinical models faithfully represent patient tumors in multiple aspects.