- Add specific copyright (2004-2025) and version (v6.0+) info for

  MSigDB Hallmark gene sets under CC BY 4.0.
- Recommend Apache 2.0 backbone alternatives for commercial users
  to ensure license compatibility.

Author: BenjaminIsaac0111

LICENSE

@@ -13,6 +13,10 @@ Use of this software, its source code, or derivative works by for-profit entitie
 
 Foundation Model Disclaimer: This software is designed to be backbone-agnostic and relies on third-party foundation models for feature extraction. A commercial agreement with the Author covers ONLY the SpatialTranscriptFormer source code, architecture, and intellectual property. It DOES NOT grant commercial use rights for restricted third-party models (e.g., CTransPath or Phikon). To use this software commercially, you must pair it with a backbone model that has a compatible commercial-friendly license (e.g., Apache 2.0 models like Hibou, Virchow, or H-Optimus-0).
 
+Dataset Disclaimer: This project utilizes third-party datasets, primarily HEST-1k. These datasets are subject to their own respective licenses (e.g., CC BY-NC-SA 4.0 for HEST-1k). A commercial agreement with the Author DOES NOT grant any rights to use, redistribute, or commercially exploit restricted third-party data. Commercial users must ensure they have independent legal rights to any training or evaluation data used within this framework.
+
+MSigDB Disclaimer: The Pathway Bottleneck and Reconstruction components utilize MSigDB Hallmark gene sets (Broad Institute). Content from MSigDB (v6.0-v7.5.1, v2022.1+) is protected by copyright (c) 2004-2025 Broad Institute, Inc., MIT, and Regents of the University of California, and is provided under CC BY 4.0. Users are responsible for complying with attribution requirements and any third-party restrictions (e.g., KEGG, BioCarta) that may apply to specific subsets of the data.
+
 Interested parties should contact the Author directly to discuss commercial licensing terms.
 
 1. ATTRIBUTION

README.md

@@ -21,6 +21,7 @@ This project is protected by a **Proprietary Source Code License**. See the [LIC
 - 🤝 **For-Profit Use**: Permitted only under a **negotiated agreement** with the author.
   - **Note on Foundation Models**: This architecture is backbone-agnostic. Any negotiated commercial agreement covers *only* the SpatialTranscriptFormer source code and IP. It does **not** grant commercial rights to use restricted third-party weights (e.g., CTransPath, Phikon). To use this system commercially, you must select a foundation model with a compatible open or commercial license (e.g., Hibou, Virchow, or H-Optimus-0).
   - **Note on HEST-1k Dataset**: The benchmark data used in this project is sourced from the **HEST-1k** dataset (Mahmood Lab), which is licensed under **CC BY-NC-SA 4.0**. This data is strictly for non-commercial research and cannot be used for commercial training or clinical deployment without explicit permission from the original authors.
+  - **Note on MSigDB**: This project uses data from the **Molecular Signatures Database (MSigDB)** (versions v6.0–v7.5.1, and v2022.1+). The contents are protected by copyright © 2004–2025 Broad Institute, Inc., MIT, and Regents of the University of California, and are distributed under the **CC BY 4.0** license. While this allows for commercial use, users must provide appropriate attribution. Note that individual gene sets within MSigDB may be subject to additional terms from third-party sources (e.g., KEGG).
 - ❌ **Prohibited**: Redistribution and unauthorized commercial exploitation.
 
 ## Intellectual Property

docs/PATHWAY_MAPPING.md

@@ -7,6 +7,8 @@ Mapping predicted gene expression into biological pathways is key to making the
 We recommend using the following curated databases for mapping:
 
 - **MSigDB Hallmark**: 50 gene sets that summarize specific biological states or processes. This is the "gold standard" for general cancer research because it's non-redundant and well-defined.
+  - **License**: MSigDB Hallmark sets (v6.0–v7.5.1, v2022.1+) are subject to the **CC BY 4.0** license.
+  - **Copyright**: © 2004–2025 Broad Institute, Inc., MIT, and Regents of the University of California.
 - **KEGG & Reactome**: More detailed, hierarchical pathways that describe specific biochemical reactions.
 - **Gene Ontology (GO)**: Useful for finding genes associated with specific molecular functions or cellular components.