MUSE-XAE

MUSE-XAE extracts mutational signatures from cancer genomes using an explainable autoencoder to enhance classification of primary tumour types and subtypes.

Key Features:

• Explainable Autoencoder Architecture: Employs a hybrid architecture with a nonlinear encoder to capture nonlinear interactions among features and a linear decoder to ensure interpretability of active signatures.
• Superior Performance in Signature Recovery: Achieves higher precision and sensitivity in recovering mutational signature profiles compared to other available tools on both synthetic and real cancer datasets.
• Tumour Classification Enhancement: Extracts highly discriminative mutational signature profiles that improve classification of primary tumour types and subtypes in real world settings.

Scientific Applications:

• Cancer Genomics Research: Facilitates understanding of genetic alterations underlying cancer development and genomic changes during tumorigenesis.
• Tumour Type Classification: Enables improved diagnosis through enhanced classification of primary tumour types and subtypes based on mutational signature profiles.
• Clinical Strategy Development: Supports the development of new strategies for diagnosis, prognosis, and treatment of cancer patients through reliable identification of underlying mutational patterns.

Methodology:

MUSE-XAE uses a hybrid explainable autoencoder architecture consisting of a nonlinear encoder and a linear decoder for mutational signature extraction.