Genetic mutation and biological pathway prediction based on whole slide images in breast carcinoma using deep learning
Problem
Identifying driver genetic mutations in breast cancer normally requires costly molecular profiling, so predicting them directly from routine whole-slide images was investigated.
Why it matters
Cheap image-based mutation prediction could help guide targeted therapies without expensive genomic testing.
What the AI did
A deep-learning model was trained on whole-slide images from the Genomic Data Commons to predict gene mutations, copy number alterations and biological pathway activities.
Results
AUC 0.68-0.85 for point mutations of six genes; AUC 0.69-0.79 for copy number alteration of another six genes; pathway activities also predicted (abstract truncated).
Impact
Demonstrates that histopathology images carry partial mutation and pathway information, with moderate prediction accuracy.
Record
| Venue | npj Precision Oncology |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Hui Yan Qu, Mu Zhou, Zhennan Yan, He Wang, Vinod K. Rustgi, Shaoting Zhang, Olivier Gevaert, Dimitris Metaxas |
| Citations | 117 (OpenAlex, at import time) |
| References | 54 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41698-021-00225-9 |
| Open access | gold |
| Keywords | deep learning, whole-slide images, breast cancer, mutation prediction, pathway activity, AUC |
Importance rating
significant — thousands to millions affected
Abstract states breast carcinoma is the most common cancer among women worldwide; cheap image-based mutation prediction could guide targeted therapies without costly genomic testing, though accuracy is moderate.