{
  "fetched_at": "2026-10-24T18:18:46Z",
  "source_url": "https://www.kaggle.com/competitions/2024-1-cuk-gm-project/data",
  "pages": [
    {
      "name": "data-description",
      "content": "## Files\n\n*   **BBBP.csv** - the Binary labels of blood-brain barrier penetration(permeability) dataset.\n*   **BBBP_train.csv**  - the train dataset.\n*   **BBBP_test.csv** - the test dataset for task 2.\n\n## Columns\n\n*   `name` - molecule name.\n*   `p_np` - blood-brain barrier or not label.\n*  `smiles`- molecule structure string.\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Dataset",
      "content": "The blood-brain barrier (BBB) is a membrane separating the blood from the brain's extracellular fluid, hence blocking out most drugs (molecules) from reaching the brain. Because of this, the BBBP has been important to study for the development of new drugs that aim to target the central nervous system. The labels for this data set are binary (1 or 0) and indicate the permeability of the molecules.\n\nThe dataset contains 2,050 molecules. Each molecule has a molecule name, a label classification and a smiles molecule string.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "Please check the Overview page.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "abstract",
      "content": "This competition is for the term project of the Graph Mining (06837-01) class at the Catholic University of Korea (CUK) in Spring 2024.\n\n# Background\nThe blood-brain barrier (BBB) is a highly selective membrane system that separates the bloodstream from the central nervous system (CNS). It plays a vital role in maintaining brain homeostasis by controlling the passage of substances between the blood and the brain tissue. Only essential nutrients and specific molecules can readily cross the BBB, while large molecules, ions, and most pathogens are excluded [1].\n\nDeveloping new drugs for neurological disorders is a challenging process, and a key hurdle is ensuring the drug can effectively penetrate the BBB. Traditionally, in-vitro and in-vivo experiments are used to assess BBB permeability, which are expensive and time-consuming. In recent years, machine learning (ML) has emerged as a powerful tool to accelerate drug discovery, and graph classification using Artificial Intelligence (AI) offers a promising approach for predicting BBB permeability of compounds.\n\nMolecules can be represented as graphs [2], where atoms are nodes and chemical bonds are edges. This representation captures the complex structure and connectivity of the molecule, which are crucial factors influencing BBB permeability. Graph classification techniques can then analyze these graphs and learn the underlying patterns that differentiate BBB-penetrating compounds from those that cannot.\n\n## References\n[1] Ballabh P, Braun A, Nedergaard M (June 2004). \"The blood-brain barrier: an overview: structure, regulation, and clinical implications\"\n[2] Baek, J., Kang, M., & Hwang, S. J. (2021). Accurate learning of graph representations with graph multiset pooling. ICLR 2021.\n\n# **Goal** \nGraph Classification using Artificial Intelligence (AI) techniques (Machine Learning, Deep Learning, Graph Neural Networks) to classify whether a compound can penetrate the blood-brain barrier (BBB) or not.\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Task",
      "content": "1. **Graph Construction:**\n  - Given the molecular strings (under SMILES format), We have to construct graphs for each SMILES string. \n  - Then, you can represent as a graph for each object, which means that 2500 molecules object will have 2500 graphs.\n\n2. **Graph Similarity**\n  - Given a few molecular strings (test dataset), finding the similar graphs.\n  - Students can use some similarity measurements, i.e., Graph Edit Distance, average_shortest_path_length, diameter, etc.)\n\n3. **Graph Classification**\n  1. Using some conventional models, i.e., SVM classifier to classify the molecules.\n    - You need to split the dataset into train/test set or train/validation/test set based on ratio. Or you can use the train/test dataset, which is already prepared.\n    - Note: Graph-based features could be degree distribution, clustering coefficient, or centrality measures.\n  2. Graph visualization \n    - Students can use T-Sne/PCA.\n\n4. **Write a report about what you found, and explain the idea, method, and result.**\n\n- Suggestion library for AI:\n  - Sklearn for Machine Learning model.\n  - Deep Learning model: Pytorch.\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "# Submission\n- Submit the code you used to verify the results and the report.\n\n# Report\n- Explain how the graph was built, what technology was used, and why it was used.\n- Explain the classification result: what technology was used, why you chose it, what you were finding, and why the result is high or low. \n- Explain the visualization: what do you think about the relationship between those graphs?\n\n## Note\n- There is no correct way to build a graph, and it is okay to use methods not learned in class to improve performance.\nOr you can use our recommendation (SMILES method from RDKit). Same for other tasks.\n- However, the report must be able to explain how the graph was built, what technology was used, and why it was used. \n- Even if the ranking is low, if the explanation in the report is understandable, you can receive a high score. \n- Conversely, even if the ranking is high, you can receive a low score if the explanation is not sufficient. \n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Rule",
      "content": "- You can use any method you think is good for the project, don't be limited by your imagination.\n- You should do it by yourself.\n- If you use another student's idea, both people will receive an F.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Contact",
      "content": "* Instructor\n  * [O-Joun Lee](https://nslab-cuk.github.io/member) - ojlee AT catholic.ac.kr\n* Teaching Assistants\n  * [Van Thuy Hoang](https://nslab-cuk.github.io/member/hoangvanthuy90) - hoangvanthuy90 AT gmail.com\n  * [Tien-Bach-Thanh Do](https://nslab-cuk.github.io/member/osfa19730) - osfa19730 AT catholic.ac.kr\n  * [Huy Quang Tran](https://nslab-cuk.github.io/member/huy) - huytran1126 AT gmail.com",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}