RareNet: a deep learning model for rare cancer diagnosis

Abstract Although significant advances have been made in the early detection of many cancers, challenges remain in the early diagnosis of rare cancers, including Wilms tumor, Clear Cell Sarcoma of the Kidney, Neuroblastoma, Osteosarcoma, and Acute Myeloid Leukemia, perhaps due to their relative obsc...

Full description

Saved in:
Bibliographic Details
Main Authors: Danyang Shao, Sohan Addagudi, Joseph Cowles, Arnav Jain, Lauren D’Souza, Steven Gore, Rajeev K. Azad
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:Scientific Reports
Online Access:https://doi.org/10.1038/s41598-025-08829-y
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849238441000173568
author Danyang Shao
Sohan Addagudi
Joseph Cowles
Arnav Jain
Lauren D’Souza
Steven Gore
Rajeev K. Azad
author_facet Danyang Shao
Sohan Addagudi
Joseph Cowles
Arnav Jain
Lauren D’Souza
Steven Gore
Rajeev K. Azad
author_sort Danyang Shao
collection DOAJ
description Abstract Although significant advances have been made in the early detection of many cancers, challenges remain in the early diagnosis of rare cancers, including Wilms tumor, Clear Cell Sarcoma of the Kidney, Neuroblastoma, Osteosarcoma, and Acute Myeloid Leukemia, perhaps due to their relative obscurity and scarce data compared to common cancers. Application of artificial intelligence or deep learning has shown promising results in disease diagnosis including in their ability to diagnose cancers and detect their tissue of origin. However, their ability to detect rare cancers is yet to be comprehensively assessed. This motivated us to develop, RareNet, which leverages transfer learning of an established deep learning model, namely, CancerNet, to classify rare cancers. The transfer learning framework of RareNet utilized DNA methylation data of various biopsied rare cancers to learn epigenetic signatures of rare cancers. RareNet achieved an overall accuracy (F1 score) of ~ 96%, outperforming other machine learning models including Random Forest, K Nearest Neighbors, Decision Tree Classifier, and Support Vector Classifier.
format Article
id doaj-art-13ff4b10f20c4d18a6d921cceae4a47c
institution Kabale University
issn 2045-2322
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-13ff4b10f20c4d18a6d921cceae4a47c2025-08-20T04:01:36ZengNature PortfolioScientific Reports2045-23222025-07-011511810.1038/s41598-025-08829-yRareNet: a deep learning model for rare cancer diagnosisDanyang Shao0Sohan Addagudi1Joseph Cowles2Arnav Jain3Lauren D’Souza4Steven Gore5Rajeev K. Azad6Department of Biological Sciences and BioDiscovery Institute, University of North TexasDepartment of Computer Sciences, University of MinnesotaDepartment of Computer Sciences, Texas Tech UniversityDepartment of Electrical Engineering and Computer Science, University of KansasDepartment of Electrical Engineering and Computer Science, University of KansasDepartment of Biological Sciences and BioDiscovery Institute, University of North TexasDepartment of Biological Sciences and BioDiscovery Institute, University of North TexasAbstract Although significant advances have been made in the early detection of many cancers, challenges remain in the early diagnosis of rare cancers, including Wilms tumor, Clear Cell Sarcoma of the Kidney, Neuroblastoma, Osteosarcoma, and Acute Myeloid Leukemia, perhaps due to their relative obscurity and scarce data compared to common cancers. Application of artificial intelligence or deep learning has shown promising results in disease diagnosis including in their ability to diagnose cancers and detect their tissue of origin. However, their ability to detect rare cancers is yet to be comprehensively assessed. This motivated us to develop, RareNet, which leverages transfer learning of an established deep learning model, namely, CancerNet, to classify rare cancers. The transfer learning framework of RareNet utilized DNA methylation data of various biopsied rare cancers to learn epigenetic signatures of rare cancers. RareNet achieved an overall accuracy (F1 score) of ~ 96%, outperforming other machine learning models including Random Forest, K Nearest Neighbors, Decision Tree Classifier, and Support Vector Classifier.https://doi.org/10.1038/s41598-025-08829-y
spellingShingle Danyang Shao
Sohan Addagudi
Joseph Cowles
Arnav Jain
Lauren D’Souza
Steven Gore
Rajeev K. Azad
RareNet: a deep learning model for rare cancer diagnosis
Scientific Reports
title RareNet: a deep learning model for rare cancer diagnosis
title_full RareNet: a deep learning model for rare cancer diagnosis
title_fullStr RareNet: a deep learning model for rare cancer diagnosis
title_full_unstemmed RareNet: a deep learning model for rare cancer diagnosis
title_short RareNet: a deep learning model for rare cancer diagnosis
title_sort rarenet a deep learning model for rare cancer diagnosis
url https://doi.org/10.1038/s41598-025-08829-y
work_keys_str_mv AT danyangshao rarenetadeeplearningmodelforrarecancerdiagnosis
AT sohanaddagudi rarenetadeeplearningmodelforrarecancerdiagnosis
AT josephcowles rarenetadeeplearningmodelforrarecancerdiagnosis
AT arnavjain rarenetadeeplearningmodelforrarecancerdiagnosis
AT laurendsouza rarenetadeeplearningmodelforrarecancerdiagnosis
AT stevengore rarenetadeeplearningmodelforrarecancerdiagnosis
AT rajeevkazad rarenetadeeplearningmodelforrarecancerdiagnosis