Study overview. Credit: Nature (2024). DOI: 10.1038/s41586-024-08167-5
A research team from Memorial Sloan Kettering Cancer Center (MSK) is demonstrating that cancer outcome predictions can be improved by breaking down hospitals’ traditional data silos and analyzing the information—including physicians’ clinical notes—with the help of artificial intelligence (AI).
A new study describes a real-time, automated approach developed at MSK that brings together doctors’ free-text notes, clinical treatment and outcomes data, patient demographic data, and tumor genomic data from the MSK-IMPACT platform to identify biomarkers that can predict outcomes and likely responses to therapy. Dubbed MSK-CHORD (for Clinicogenomic Harmonized Oncologic Real-World Dataset), the effort is the largest of its kind, combing data from nearly 25,000 patients with non-small cell lung, breast, colorectal, prostate, and pancreatic cancers.
The study was led by co-first authors Justin Jee, MD, Ph.D., Christopher Fong, Ph.D., Karl Pichotta, Ph.D., Thinh Ngoc Tran, Ph.D., and Anisha Luthra, and overseen by senior author Nikolaus Schultz, Ph.D., Director of MSK’s Cancer Data Science Initiative. It is published in the journal Nature.
The team found that cancer outcome predictions based on MSK-CHORD data outperformed those based on genomic data or cancer stage alone. By analyzing more than 700,000 radiology reports, MSK-CHORD was able to uncover predictors of metastasis to specific organ sites. Additionally, MSK-CHORD’s size and rich annotations led the team to identify mutations in the SETD2 gene as an uncommon but promising biomarker of immunotherapy response in lung adenocarcinoma—a finding that was corroborated in multiple independent datasets.
“Our results highlight the power of natural language processing and the impact of bringing together a multitude of data streams to better predict patient outcomes,” says Dr. Jee, a thoracic medical oncologist at MSK.
“It is our hope that MSK-CHORD will fuel further research into the relationships between genomic data and real-world outcomes in cancer,” Dr. Schultz adds.
Dozens of clinicians and researchers from across MSK came together to share expertise, and to develop the natural language processing models, AI risk models, and engineering infrastructure required to support the effort, they note.
More information:
Justin Jee et al, Automated real-world data integration improves cancer outcome prediction, Nature (2024). DOI: 10.1038/s41586-024-08167-5
Provided by
Memorial Sloan Kettering Cancer Center
Citation:
Automated approach breaks down data silos to better predict cancer outcomes (2024, November 6)
retrieved 7 November 2024
from https://medicalxpress.com/news/2024-11-automated-approach-silos-cancer-outcomes.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no
part may be reproduced without the written permission. The content is provided for information purposes only.
