init commit of readme

README.md

@@ -1,3 +1,195 @@
----
-license: mit
----
+---
+annotations_creators: []
+language:
+- en
+license: other   # DepMap Terms of Use (non-commercial, research-only)
+multilinguality: []
+pretty_name: Target-QA (DepMap-based Multi-Omic & Knowledge Graph Dataset)
+size_categories:
+- 1K<n<10K
+source_datasets:
+- DepMap (Broad Institute of MIT and Harvard)
+- Cellosaurus
+task_categories:
+- question-answering
+- text-retrieval
+- graph-ml
+- other
+task_ids:
+- open-domain-qa
+- document-retrieval
+paperswithcode_id: galax-target-qa
+---
+
+
+
+# Target-QA: QA Dataset Based on DepMap Multi-Omics Profiles and CRISPR Outcomes
+
+## Dataset Summary  
+
+**Target-QA** is a dataset constructed from the **DepMap** multi-omics and CRISPR screening cohorts, harmonized via **BioMedGraphica**.  
+It enables **multi-modal reasoning** by combining quantitative omics, biomedical knowledge graphs, and disease annotations for **CRISPR target prioritization**.  
+
+The dataset is designed for training and benchmarking **graph-augmented large language models (LLMs)**, such as **GALAX**, that reason over both structured and unstructured biological information.  
+
+---
+
+## Supported Tasks and Benchmarks  
+
+- **Graph-Augmented Question Answering**  
+  - Identify CRISPR targets given omics and KG context.  
+- **Knowledge Graph Reasoning**  
+  - Subgraph extraction and signaling network prioritization.  
+- **Multi-Omic Target Prioritization**  
+  - Integration of genomic, transcriptomic, and proteomic data.   
+
+---
+
+## Dataset Structure  
+
+### Example JSON  
+
+```json
+{
+  "cell_line_name": "GAMG",
+  "cell_line_id": "ACH-000098",
+  "disease": "glioblastoma",
+  "disease_bmgc_id": "BMGC_DS00965",
+  "sample_dti_index": 123,
+  "input": {
+    "top_k_gene": {
+      "hgnc_symbols": ["EGFR", "CDKN2A"],
+      "protein_bmgc_ids": ["BMGC_PR01234"],
+      "protein_llmname_ids": ["ENSP00000354587"]
+    },
+    "top_k_transcript": {...},
+    "top_k_protein": {...},
+    "knowledge_graph": {
+      "disease_protein": {...},
+      "ppi_neighbors": {...},
+      "protein_relationships": ["BRCA1 → TP53"]
+    }
+  },
+  "ground_truth_answer": {
+    "hgnc_symbols": ["TP53", "EGFR"],
+    "protein_bmgc_ids": ["BMGC_PR00987", "BMGC_PR04567"],
+    "protein_llmname_ids": ["ENSP00000439978"]
+  }
+}
+
+```
+
+## Data Fields  
+
+- **cell_line_name**: Cell line name *(string)*  
+- **cell_line_id**: DepMap identifier *(string)*  
+- **disease**: Disease name *(string)*  
+- **disease_bmgc_id**: BioMedGraphica disease ID *(string)*  
+- **sample_dti_index**: Omics index for NumPy array access *(int)*  
+- **input**: Multi-modal context  
+  - Top-k genes, transcripts, proteins (HGNC, BioMedGraphica IDs, synonyms)  
+  - Knowledge graph neighbors, PPIs, disease–protein associations  
+- **ground_truth_answer**: CRISPR-validated targets (HGNC, BioMedGraphica IDs, synonyms)  
+
+---
+
+## Preprocessing Details  
+
+![Dataset composition and splits](FigureS1.pdf)
+
+- **Starting cohort**: 985 DepMap cell lines  
+  - 649 annotated (cancerous)  
+  - 336 non-annotated or non-cancerous  
+- **Target-QA subset**: 363 overlapping with CRISPR gene effect data  
+- **Omics modalities integrated into 834,809 entities**:  
+  - Promoter: 86,238  
+  - Gene: 86,238  
+  - Transcript: 412,039  
+  - Protein: 121,419  
+- **Knowledge graph integration**:  
+  - Protein–protein interactions: 17,151,453 edges  
+  - Disease–target associations: 27,087,971 edges  
+
+> **Note:** Methylation values excluded due to high saturation.  
+
+---
+
+## Data Splits  
+
+- **Pretraining set**: 336 samples  
+  - Train: 269  
+  - Test: 67  
+- **Target-QA set**: 363 samples  
+  - Train: 300  
+  - Test: 63  
+
+**Test distribution includes:** LUAD (7), BRCA (6), COAD/READ (5), PAAD (4), GBM (3), SARC (3), OV (3), SKCM (3), ESCA (3), SCLC (3), HNSC (2), LUSC (2), STAD (2), etc.  
+
+---
+
+## Prompt Design  
+
+### Initial Prompt ($P^{init}_n$)  
+
+**Inputs:**  
+- Top-10 ranked genes, transcripts, proteins  
+- Disease-associated proteins from KG  
+- Known PPI and disease–protein relationships  
+
+**Output:**  
+- 100 vulnerability genes $r^{(init)}_{n,1...100}$  
+
+### Refined Prompt ($P^{final}_n$)  
+
+**Inputs:**  
+- Same as above  
+- + Subsignaling gene regulatory network (from graph generator)  
+
+**Output:**  
+- Refined 100 vulnerability genes $\hat{r}_{n,1...100}$  
+
+---
+
+## Intended Uses  
+
+- Pretraining and fine-tuning **Graph-Language Foundation Models (GLFMs)**  
+- **Target prioritization** tasks in cancer research  
+- **Benchmarking interpretable biomedical AI**  
+- **Graph-based multi-modal reasoning** in bioinformatics  
+
+---
+
+## Limitations  
+
+- Not suitable for clinical or diagnostic use  
+- Limited to DepMap cell line coverage (not all cancer types)  
+- Excludes methylation modality  
+
+---
+
+## Licensing  
+
+This dataset is based on **DepMap** data and is subject to the **DepMap Terms of Use**:  
+- Free for **research purposes only**  
+- **Commercial use prohibited** without explicit license from Broad Institute or contributors  
+- Machine learning models may be trained **for internal research use** or shared **for non-profit research purposes**  
+- Attribution to **Broad Institute / DepMap** is required  
+
+**DepMap Terms of Use:** [https://depmap.org/portal/termsOfUse](https://depmap.org/portal/termsOfUse)  
+
+**Suggested acknowledgment:**  
+> "The results here are in whole or part based upon data generated by the DepMap, Broad Institute of MIT and Harvard."  
+
+---
+
+## Citation  
+
+If you use this dataset, please cite:  
+
+```bibtex
+@article{zhang2025galax,
+  title={GALAX: Graph-Augmented Language Model with Explainability for CRISPR Target Prioritization},
+  author={Zhang, Heming and Li, Fuhai and Chen, Yixin and others},
+  year={2025},
+  journal={Preprint}
+}