Unveiling cancer risk through AI-powered spatial insights

The National Cancer Institute awarded the Institute For Systems Biology $555,592 on September 1, 2026, under the Cancer Cause and Prevention Research program (CFDA 93.393) to develop AI-driven infrastructure for integrating and analyzing cancer datasets and identifying cancer risk factors.

The recipient will construct a multi-modal, harmonized resource for the tumor microenvironment featuring an AI-powered spatial search tool. Using spatial datasets from the Human Tumor Atlas Network (HTAN), the project will extract cellular and molecular features, generate digital pathology embeddings through deep learning, and create a searchable database for tissue similarity analysis. This database will enable researchers to build multi-center cohorts and explore cancer's spatial landscape. The initiative also addresses the need for streamlined analysis tools to extract meaningful insights from rapidly expanding genomic, transcriptomic, proteomic, and metabolomic datasets by making advanced analytical methodologies including machine learning more accessible.

The award runs through August 31, 2031, with place of performance in Seattle, Washington. The assistance type is a Project Grant.

Abstract

Cancer research has greatly benefited from large-scale data-sharing efforts by governments, universities, and research institutions, leading to valuable datasets that have advanced our understanding of tumor biology. One of the most significant initiatives, the cancer genome atlas (tcga), has generated and shared genomic data across 30 tumor types. Alongside tcga, many other data-sharing projects continue to expand, providing increasingly detailed insights into tumor genomics, transcriptomics, proteomics, and metabolomics at the bulk, single-cell, and spatial levels. These advances have enabled researchers to study the tumor-immune microenvironment with unprecedented resolution, but the rapid expansion of data has created a need for streamlined analysis tools to extract meaningful insights about cancer risk, progression, and treatment response. To address this challenge, this initiative proposes to build an AI-driven infrastructure to improve the integration and analysis of cancer datasets. This system will combine data sources with strengths in clinical annotations, genomic depth, and spatial analysis, making advanced analytical methodologies like machine learning more accessible. A key component of this initiative is the development of a large-scale histopathology database, applying deep learning to enhance cross-consortium data analysis. The project has three main aims. First, researchers will construct a multi-modal, harmonized resource for the tumor microenvironment, including an AI-powered spatial search tool. By leveraging spatial datasets from htan, they will extract cellular and molecular features, generate digital pathology embeddings, and create a searchable database for tissue similarity analysis. This will allow researchers to build multi-center cohorts and explore cancer's spatial landscape in greater depth. Second, they will identify and characterize factors associated with cancer risk by integrating data from major resources such as CRI iatlas, aacr project genie, cbioportal, and htan. This analysis will link molecular and cellular features to patient outcomes, immunotherapy response, and tumor environments, helping to uncover mechanisms underlying cancer susceptibility and treatment effectiveness. Lastly, the project will map spatial structures, immune pathways, and cellular networks linked to disease progression and therapy response. By analyzing tumor samples at different disease stages, researchers aim to understand how cancer evolves and disrupts immune function, providing insights that could lead to better therapeutic strategies. Ultimately, this initiative will enhance cancer risk assessment, refine predictive models, and improve precision medicine approaches through the integration of patient outcomes data with deep spatial multi-omics using AI- driven analysis.