Skip to contentSkip to filters

We use cookies for analytics.

Seminar, Hosted by Flatiron Institute

AstroPath: Astronomy Meets Pathology to Study Cancer Immunotherapy

Recording

1 h 2 min

Abstract

This talk describes a detailed, automated, multi-step approach to multispectral staining and image analysis of multiplex immunofluorescent (mIF) slides of cancerous tissue. It supports the study of cancer immunotherapy and the complex tumor microenvironment (TME) by quantifying co-expression patterns of key markers on specific cells and spatial relationships between interacting cell types.

The AstroPath pipelines have processed more than 20 cohorts of 1,000 slides into the AstroPath database, totalling 60,000 high-powered fields and 1.5 billion detected cells, of which a unique statistical sample of 1 billion distinct cells has full geometric boundaries for nuclei and membranes. To support TME studies, the databases hold precomputed neighbors within 25 microns of each cell, 3.5 billion cell pairs with their phenotypes, derived from image processing over 65 trillion pixels; by comparison, the Sloan Digital Sky Survey collected about 6.4 trillion raw pixels over 16 years of imaging. The group brings together astronomers, particle physicists, machine-vision and deep-learning experts and cancer pathologists, systematically adapting lessons from sky surveys to cancer immunotherapy; the resulting platform shows the feasibility of an automated system feeding an open public database, with open-source tumor-immune atlases of billions of spatially mapped cells to follow.

Using AstroPath, the team discovered in 2021 the first predictive biomarker for melanoma, which predicts with high significance whether a patient will respond to current immunotherapy drugs. AI is increasingly used both for code development and for testing hypotheses about new pathways in cancer-immune interactions, and a large multimodal foundation model is being built that incorporates genomics data from The Cancer Genome Atlas (TCGA) and integrates the melanoma collection with spatial imaging from the same patients. The talk describes how lessons from astronomy short-circuited the design and implementation process.

Topics