Deep learning improves tumor classification by detecting unknown tumor types
Enhancing Clinical Classifier Robustness with Deep Representation Learning Based OOD Detection on DNA Methylation Data
Artificial IntelligenceMachine Learning
Summary
Tumor classification using DNA methylation data helps doctors diagnose cancer types, but existing computer models can make mistakes when they encounter tumor types they were not trained on. The authors developed a new deep learning method that can recognize when a sample is unlike those seen before, improving the model's safety and reliability. They tested several techniques and found that a method called 1D-Subspace Analysis worked best to identify unknown tumors. This approach can serve as a quality control layer, making tumor classification tools more trustworthy in clinical settings.
What this means in practice
- •For clinical diagnostic teams: Add a deep learning based quality check to DNA methylation classifiers to detect tumor types not seen during training and reduce misdiagnosis risk.
- •For clinical bioinformatics teams: Integrate out-of-distribution detection methods into tumor analysis pipelines to flag samples with unexpected methylation patterns for further review.
Authors
Md Zahangir Alom, Quynh T. Tran, Breuer Alexandar, Brent A. Orr
Abstract
DNA methylation (DNAM) array-based tumor classification is a powerful clinical diagnostic tool, but models trained on fixed reference cohorts can produce faulty predictions for tumor types not represented in the training distribution. We propose an end-to-end deep representation learning-based out-of-distribution (OOD) detection framework for DNAm array data. Deep neural network (DNN) classifiers were trained on public sarcoma (SARC) (GSE140686) and brain tumor (BT) (GSE90496) cohorts, and feature vectors were extracted from bottleneck layers. Five OOD methods, including 1dimension (1D)-Subspace Analysis (1D-SSA), K-Nearest Neighbors (KNN), Ridge Regression, Mahalanobis Distance (MD), and Sum of Squared Distances (SSD), were explored with 200 and 500-dimensional embeddings. Spectral discrepancy was used to discriminate in-distribution (IND) from OOD samples. Performance was measured via cross-validation, holdout samples, and an independent clinical dataset. 1D-SSA was the most effective approach. In the sarcoma-IND/brain tumor-OOD setting, the framework achieved an AUROC of 0.975 and True Negative Rate (TNR)@95TPR of 0.930, and an AUROC of 0.9569 and TNR@95TPR of 0.8660 for the brain tumor-IND/sarcoma-OOD setting. On the independent clinical test set, the method yielded a mean AUROC of 0.976, a DE of 0.032, an FPR@95TPR of 0.065, and a TNR@95TPR of 0.935. These results show deep representation-based OOD detection as a practical quality-control layer for clinical-grade methylation-based tumor classification systems.