K-NeSyNet: Experimental Source Data and Reproducibility Package for Personalized Oncology Treatment Recommendation

Résumé fourni par la source

This dataset and reproducibility package supports the study “Knowledge-Driven Neuro-Symbolic Reasoning for Personalized Oncology Treatment Recommendation Using a Multimodal Medical Knowledge Graph (K-NeSyNet).” It contains experimental source data for the main benchmark comparison, ablation analyses, safety-loss sensitivity analysis, hyperparameter sensitivity analysis, cancer-specific modality analysis, target-channel coverage analysis, and case-level score decomposition. The package also provides cohort-construction summaries, knowledge-graph leakage audit results, clinical-text filtering statistics, treatment-label distribution summaries, source and configuration records, and reproducibility utilities supporting the reported evaluation protocol. The final analytic cohort comprises 4,781 TCGA cases from 10 cancer projects, divided into 3,346 training, 717 validation, and 718 test cases. Original TCGA/GDC source files are not redistributed in this archive and remain available through the NCI Genomic Data Commons under the corresponding open- or controlled-access requirements. Third-party biomedical knowledge resources remain subject to their respective licensing and reuse conditions.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.