Systems Pharmacology of Therapeutic and Adverse Responses to ImmuneCheckpoint and Small Molecule Drugs

点击翻译按钮获取中文摘要

SUMMARY ABSTRACT Single cell RNA profiling and DNA sequencing has revolutionized our understanding of the tumor microenvironment (TME) but such data lacks the spatial context and morphological information found in images. Histology makes extensive use of morphology, and in a clinical setting provides the primary means of diagnosing disease and managing treatment. However, relatively little molecular insight can be obtained from classical Hematoxylin and Eosin (H&E) or immunohistochemistry. These considerations have led to the recent development of multiple methods for performing highly multiplexed tissue imaging. It allows the properties of single cells to be determined in a preserved 3D environment in humans and mouse models. In a research setting, multiplexed imaging provides new insight into molecular mechanisms of tumor initiation, progression, immune editing, and escape. In a clinical setting, high-plex imaging promises to augment the traditional histopathological diagnosis of disease with molecular information needed to guide use of targeted and immuno-therapies. Multiple high-plex tissue images yield subcellular resolution data on 20-100 proteins or other biomolecules on resolvable structures having spatial scales from 100 nm to over 1 cm in specimens as large as 5 cm2. These images contain 106 to 107 cells, encoded in up to 1 TB of data. The primary barrier to wider use of high-plex imaging centers on the computational challenges associated with processing, managing, and disseminating images of this size. Many algorithms and methods have been developed to process images of cells grown in culture and these provide a foundation for analysis of tissue images. However, high-plex tissue imaging poses many additional challenges arising from the diversity and crowding of cells and as well as the size of the data. We have constructed a cloud-deployed pipeline (MCMICRO) that uses Docker-containers and a NextFlow pipeline to process large-scale tissue images and generate single-cell data in a standardized format. We propose to reengineering the components of this proof- of-concept implementation to make it performative and broadly useful. Aim 1 will improve the performance of individual modules through code profiling and optimization. Aim 2 will complete the general user and programmer documentation of MCICRO and its modules to enable continued contributions from the open- source community and to increase interoperability and perform testing. Aim 3 will add modules to MCMICRO based on existing proof-of concept code available in the public domain. Aim 4 will enable output of pipeline intermediate and final results - including image data itself - from the cloud without requiring download. These supplementary aims are directly relevant the approved aims of the parent award. Completing these aims will involve partial reengineering of existing software modules and evaluation of pipeline performance using real- world test data that we will release as part of this supplement. All Aims involve executing MCMICRO on Amazon and Google cloud platforms.