neoDL

neoDL predicts survival and stratifies IDH wild-type glioblastomas by analyzing neoantigen intrinsic features with a deep learning model to inform prognostic subgrouping.

Key Features:

• Neoantigen intrinsic feature-based modeling: Utilizes intrinsic molecular features of neoantigens associated with patient survival outcomes.
• Deep learning framework: Implements a deep learning architecture to predict survival with reported AUC 0.988 and p-value <0.0001 when applied to the TCGA dataset.
• Prognostic subgroup stratification: Stratifies IDH wild-type GBMs into distinct prognostic subgroups validated by leave-one-out cross-validation (LOOCV) on the TCGA cohort and an independent Asian population dataset.
• Protective feature identification: Identifies 12 protective neoantigen intrinsic features enriched in pathways related to development and cell cycle associated with long-term survival.

Scientific Applications:

• Neoantigen-based personalized immunotherapy: Enables selection of IDH wild-type GBM cases with favorable neoantigen profiles for neoantigen-targeted immunotherapeutic strategies.
• Prognostic biomarker discovery: Provides candidate neoantigen intrinsic features as prognostic biomarkers for patient outcome prediction.
• Research on immunogenicity: Offers insights into neoantigen characteristics with high immunogenic potential to inform cancer immunotherapy research.

Methodology:

Deep learning model training and prediction; leave-one-out cross-validation (LOOCV) on TCGA; independent validation in an Asian population dataset; calculation of AUC (0.988) and statistical significance (p < 0.0001); identification of 12 protective neoantigen intrinsic features and pathway enrichment analysis indicating development and cell cycle pathways.