Masked Autoencoder Joint Learning for Robust Spitzoid Tumor Classification

2025-11-24

AI Summary

Accurate molecular subtyping of spitzoid tumors (STs) is critical for prognosis assessment and therapeutic decision-making; however, clinical DNA methylation data frequently suffer from high-dimensional missingness due to insufficient sequencing coverage and technical artifacts, severely limiting the performance of existing classification models. To address this, we propose ReMAC—the first masked autoencoder framework specifically designed for tumor methylation analysis. ReMAC unifies representation learning, missing-value imputation, and end-to-end classification within a single architecture, enabling robust modeling of both complete and incomplete methylation profiles. Notably, it introduces masked autoencoding—a paradigm previously unexplored in epigenetic classification—to jointly reconstruct masked features and optimize discriminative subtyping. Evaluated on real-world clinical ST datasets, ReMAC achieves significant improvements in subtype classification accuracy and robustness over state-of-the-art methods. The implementation is publicly available.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Accurate diagnosis of spitzoid tumors (ST) is critical to ensure a favorable prognosis and to avoid both under- and over-treatment. Epigenetic data, particularly DNA methylation, provide a valuable source of information for this task. However, prior studies assume complete data, an unrealistic setting as methylation profiles frequently contain missing entries due to limited coverage and experimental artifacts. Our work challenges these favorable scenarios and introduces ReMAC, an extension of ReMasker designed to tackle classification tasks on high-dimensional data under complete and incomplete regimes. Evaluation on real clinical data demonstrates that ReMAC achieves strong and robust performance compared to competing classification methods in the stratification of ST. Code is available: https://github.com/roshni-mahtani/ReMAC.

Problem

• Classifying spitzoid tumors using incomplete DNA methylation data
• Addressing missing entries in methylation profiles for diagnosis
• Developing robust classification methods for high-dimensional clinical data

Innovation

• Masked Autoencoder joint learning for robust classification
• Extension of ReMasker for incomplete methylation data
• Handles high-dimensional data with missing entries