# edouard-rabasse/hackaton-margo architecture

> Architecture diagram of the GitHub repository edouard-rabasse/hackaton-margo, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2025-05-20). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/edouard-rabasse/hackaton-margo
- Repository: https://github.com/edouard-rabasse/hackaton-margo

## Overview

##### Project type and purpose
- This is a 48-hour hackathon codebase for molecular toxicity prediction, organized by Margo, IBM, and Qubit Pharmaceuticals.
- It’s an end-to-end ML research & prototyping project that (a) ingests SMILES strings and molecular descriptors, (b) featurizes them, (c) trains multiple models (GNNs, DNNs on fingerprints/descriptors, XGBoost+Optuna), and (d) combines outputs for submission.

##### File structure
Top-level:
- main.py & models.py     → likely entry-point orchestrators for experiments
- environment.yml & requirements.txt
- visu.py                 → visualization of results
- ChembertA/chemberta.py  → BERT-based chemical embedding prototype
- DeepHIT/…               → core GNN + DeepHit model pipelines
- edouard/xgboost_optuna.py → XGBoost with Optuna hyperparameter tuning
- simon/…                 → Jupyter notebooks for exploratory ML pipelines

DeepHIT folder structure shows:
- train_*.py              → separate scripts for GCN, descriptor DNN, fingerprint DNN, fully-connected
- gnn.py, gnn2.py, GNN3.py → successive GNN implementations
- combine*.py             → ensemble/stacking pipelines
- utils.py                → shared functions
- main.py                 → pipeline entry
- models.py               → model definitions

##### README insights
- Project goal: predict toxicity from SMILES + molecular features.
- Approach: represent molecular geometry via Graph Neural Networks, adapt DeepHit model.

## Diagram (Mermaid)

```mermaid
flowchart TD
    %% Data & Preprocessing
    subgraph "Data & Preprocessing"
        RawData[(Raw Dataset SMILES & Labels)]:::data
        Split["train_test_split.py"]:::data
    end
    RawData --> Split

    %% Featurization
    subgraph "Featurization"
        DescExtract["Descriptor Extractor\n(utils.py)"]:::data
        Fingerprint["Fingerprint Generator\n(utils.py)"]:::data
        ChemBERTa["ChemBERTa Embedder\n(chemberta.py)"]:::data
        GraphBuilder["Graph Builder\n(gnn.py, gnn2.py, GNN3.py)"]:::data
    end
    Split --> DescExtract
    Split --> Fingerprint
    Split --> ChemBERTa
    Split --> GraphBuilder

    %% Model Training
    subgraph "Model Training"
        GNNPipeline["DeepHIT Pipeline\n(main.py)"]:::train
        DNNDesc["Descriptor DNN\n(train_dnn_descriptor.py)"]:::train
        DNNFP["Fingerprint DNN\n(train_dnn_fingerprint.py)"]:::train
        FCFc["FC-FC Baseline\n(train_fcfc.py)"]:::train
        GCN["GCN Baseline\n(train_gcn.py)"]:::train
        XGB["XGBoost+Optuna\n(xgboost_optuna.py)"]:::train
    end
    DescExtract --> DNNDesc
    Fingerprint --> DNNFP
    ChemBERTa --> GNNPipeline
    GraphBuilder --> GNNPipeline
    Split --> XGB
    GNNPipeline --> Combine
    DNNDesc --> Combine
    DNNFP --> Combine
    FCFc --> Combine
    GCN --> Combine
    XGB --> Combine

    %% Ensemble & Submission
    subgraph "Ensemble & Submission"
        Combine["combine.py & combine_pipeline.py"]:::ensemble
        Submit["submission.py"]:::ensemble
    end
    Combine --> Submit

    %% Orchestration
    subgraph "Orchestration"
        Main["main.py"]:::orch
    end
    Main --> Split
    Main --> DescExtract
    Main --> Fingerprint
    Main --> ChemBERTa
    Main --> GraphBuilder
    Main --> GNNPipeline
    Main --> DNNDesc
    Main --> DNNFP
    Main --> FCFc
    Main --> GCN
    Main --> XGB
    Main --> Combine
    Main --> Submit

    %% Visualization & Reporting
    subgraph "Visualization & Reporting"
        Vis["visu.py"]:::viz
    end
    GNNPipeline --> Vis
    Combine --> Vis

    %% Exploratory Artifacts
    subgraph "Exploratory Artifacts"
        NB1["ML_pipeline.ipynb"]:::art
        NB2["xgboost.ipynb"]:::art
        NB3["optuna_study.json"]:::art
        NB4["optuna_study.pkl"]:::art
    end
    NB1 --> XGB
    NB2 --> XGB

    %% Documentation
    subgraph "Documentation"
        MD1["DeepHIT/README.md"]:::doc
        MD2["DeepHIT/research.md"]:::doc
        MD3["README.md"]:::doc
        MD4["environment.yml\nrequirements.txt"]:::doc
    end

    %% Click Events
    click Split "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/train_test_split.py"
    click DescExtract "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/utils.py"
    click Fingerprint "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/utils.py"
    click ChemBERTa "https://github.com/edouard-rabasse/hackaton-margo/blob/main/ChembertA/chemberta.py"
    click GraphBuilder "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/gnn.py"
    click GNNPipeline "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/main.py"
    click DNNDesc "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/train_dnn_descriptor.py"
    click DNNFP "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/train_dnn_fingerpint.py"
    click FCFc "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/train_fcfc.py"
    click GCN "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/train_gcn.py"
    click XGB "https://github.com/edouard-rabasse/hackaton-margo/blob/main/edouard/xgboost_optuna.py"
    click Combine "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/combine.py"
    click Submit "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/submission.py"
    click Main "https://github.com/edouard-rabasse/hackaton-margo/blob/main/main.py"
    click NB1 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/simon/ML_pipeline.ipynb"
    click NB2 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/simon/xgboost.ipynb"
    click NB3 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/simon/optuna_study.json"
    click NB4 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/simon/optuna_study.pkl"
    click Vis "https://github.com/edouard-rabasse/hackaton-margo/blob/main/visu.py"
    click MD1 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/README.md"
    click MD2 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/DeepHIT/research.md"
    click MD3 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/README.md"
    click MD4 "https://github.com/edouard-rabasse/hackaton-margo/blob/main/environment.yml"

    %% Styles
    classDef data fill:#bbdefb,stroke:#1e88e5,color:#0d47a1;
    classDef train fill:#c8e6c9,stroke:#43a047,color:#1b5e20;
    classDef ensemble fill:#ffe0b2,stroke:#fb8c00,color:#e65100;
    classDef orch fill:#e1bee7,stroke:#8e24aa,color:#4a148c;
    classDef viz fill:#eeeeee,stroke:#757575,color:#212121;
    classDef art fill:#f5f5f5,stroke:#bdbdbd,color:#424242;
    classDef doc fill:#f3e5f5,stroke:#ab47bc,color:#6a1b9a;
```
