The new funding marks a deliberate shift from discovery to translation. During the first phase of the Human Tumor Atlas Network, Campbell and colleagues helped establish the Lung Pre-Cancer Atlas, a cellular map of the earliest stages of lung cancer development under the leadership of Avrum Spira. That atlas catalogued, at remarkable molecular resolution, the changes that unfold in lung tissue as normal cells drift toward malignancy. The Phase II award now allows the team to convert that encyclopedic map into something a physician can actually use at the bedside: biomarkers that distinguish dangerous nodules from benign ones before a surgeon’s knife is ever considered.

The clinical problem the team is attacking is both common and consequential. Lung adenocarcinoma is the most frequently diagnosed form of lung cancer and remains a leading cause of cancer death worldwide, even as smoking cessation programs, low-dose CT screening, targeted therapies and immunotherapies have improved outcomes for many patients. Screening itself has created a new challenge: as more people undergo CT scans, radiologists increasingly find so-called ground glass nodules, faint hazy areas that can represent early adenocarcinoma, precancerous change, inflammation or old scars. Telling these possibilities apart is notoriously difficult, and the stakes of guessing wrong run in both directions.

A nodule that is actually indolent may trigger an invasive surgical resection, exposing a healthy patient to the risks, pain and anxiety of a major procedure. A nodule that is quietly progressing toward invasive cancer, meanwhile, may be watched for too long, and the window for curative early intervention can close. Campbell’s team describes the goal as eliminating this guesswork through a multimodal strategy that reduces false-negative rates, ensuring earlier and more targeted intervention while minimizing overtreatment. To do so, the investigators have assembled a multidisciplinary group spanning pulmonology, pathology, cancer biology, imaging science and data science at Boston University, Roswell Park Comprehensive Cancer Center in Buffalo, the University of Colorado Anschutz, and the University of California, Los Angeles.

Central to the project is a new generation of sampling technology. Robotic-assisted bronchoscopy systems use steerable, computer-navigated instruments that can travel deep into the peripheral lung and reach small nodules that conventional bronchoscopes struggle to access. The team will systematically evaluate how well these emerging platforms perform on part-solid ground glass nodules, which are small, hazy and less dense than solid nodules and are therefore harder to reach and harder to sample. A biopsy is judged successful when it retrieves enough tissue to yield a definitive answer, whether cancer, precancer or benign, and the proportion of such definitive results is known as the diagnostic yield, the key metric for whether a biopsy platform is worth using at all. Because of the physical properties of these nodules, a substantial fraction of current biopsies come back inconclusive, leaving patients and physicians without an answer and often facing repeat procedures or surgery.

To build one of the largest cohorts of part-solid nodule biopsies assembled to date, interventional pulmonologists at Roswell Park, led by co-principal investigator Nathaniel Ivanick, MD, and at Boston Medical Center, led by co-investigator Ehab Billatos, MD, will collect biopsy specimens across the participating centers. Every sample will then be reviewed by co-principal investigator Daniel Merrick, MD, a pathologist at the University of Colorado who specializes in the earliest stages of lung cancer, ensuring that the histological ground truth against which the biomarkers are calibrated is read by an expert in precisely the lesions the project targets.

Imaging will provide a second, independent stream of evidence. William Hsu, PhD, at UCLA, will lead automated quantitative image analysis that extracts advanced radiomic features from CT scans, measuring characteristics of the nodule itself, the surrounding lung tissue and the lung as a whole. The team will determine which of these features best predict whether a given biopsy will produce a usable diagnosis. That predictive capability could change clinical practice before any molecular test is run: if a scan can forecast, in advance, which nodules are likely to yield inconclusive biopsies, physicians could choose which nodules to sample, select the most appropriate tool, and plan a route to diagnosis that requires fewer procedures overall.

The molecular layer of the project will be directed by co-principal investigator Sarah Mazzilli, PhD, assistant professor of medicine at Boston University and Director of the BU Spatial Biology Core. Biopsy material, which is inherently limited in quantity, will be profiled using spatial transcriptomics, a technology that maps which genes are active in each histological region of a tissue sample rather than averaging signals across the whole specimen. These spatially resolved molecular profiles will be integrated with CT imaging features and artificial intelligence-based risk scores to build models capable of diagnosing cancer even in biopsies that pathologists judged inconclusive, and of predicting which nodules are likely to progress toward invasive disease.

Mazzilli emphasized the precision the approach demands. Each biopsy is tiny, she noted, so every piece of tissue has to count. Her team has built a pipeline to optimally preserve and process each sample, profile it, and map gene activity across it at near single-cell resolution, allowing the earliest molecular signs of cancer to be pinpointed even when they are not yet distinguishable under the microscope. Just as important, she added, every profile generated will become part of a public atlas, so that researchers everywhere can apply their own novel methods to better understand how lung cancer begins and how it might be targeted for intervention before it becomes lethal.

The broader context for the work is the Human Tumor Atlas Network itself, a National Cancer Institute initiative to construct three-dimensional atlases of the dynamic cellular, morphological and molecular features of human cancers as they evolve from precancerous lesions to advanced disease. By anchoring its biomarker development in that atlas framework, the Boston University-led consortium hopes to do more than refine the management of a single nodule type. If the integrated models perform as intended, the project could establish a template for how large-scale molecular atlases are converted into clinical decision tools, transforming early-stage lung cancer management so that patients with aggressive tumors receive immediate treatment while healthy individuals are spared the anxiety and physical toll of unnecessary procedures. The research is supported by the National Cancer Institute of the National Institutes of Health under Award Number U01CA313003, and its content remains the responsibility of the investigators rather than the National Institutes of Health.

Subject of Research: Translating the Lung Pre-Cancer Atlas into multimodal biomarkers for managing lung ground glass nodules

Article Title: Grant awarded to BU-led team targeting early lung cancer detection

Article References: Grant awarded to BU-led team targeting early lung cancer detection. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: lung cancer, ground glass nodules, Human Tumor Atlas Network, Lung Pre-Cancer Atlas, biomarkers, robotic-assisted bronchoscopy, spatial transcriptomics, radiomics, early detection, lung adenocarcinoma, Boston University, National Cancer Institute