How humanomelab/mlcourse works
Main Components
• Data Repository Layer: Represent the primary data sources and storage. This includes the “data” folder (with subdirectories for images, sequences, and CSV data) and the “img” folder (which may hold additional visualization outputs).
• Notebook/Exploratory Environment: The Jupyter notebooks (0_data_prep_and_visualization.ipynb through 4_deep_learning_for_sequences.ipynb) serve as interactive notebooks guiding users through different steps in the machine learning workflow.
• Script Modules: The “script” folder includes standalone data processing utilities (e.g., extract_seq.py, get_chipseq_peak_info.py, random_seq_except_listed.py) that likely support or preprocess data prior to analysis.
• Model Implementation/Training Components: The “src” directory hosts Python modules (chipseq_prediction_cnn.py, yeast_feature_cnn.py, yeast_image_cnn.py) where deep learning models (CNNs) are implemented for prediction tasks on different data modalities.
Out the Data Flow and Relationships
• Feature-based Machine Learning Workflow: Follow the data flow into notebooks that perform ML using features (as in chapter 1 and 2 notebooks).
• Interconnections: Data prepared and features extracted feed into the corresponding neural network models found in the “src” folder, while additional scripts in “script” help process and tailor the input data.
• External Data Integration: Although not explicitly an API interaction, note that the README mentions resources such as SCMD and ChIP-Atlas, so a cloud/external data source component may be represented to show the origin of data or related external references.
Architectural Patterns and Technologies
• Technologies: Mention Jupyter Notebook for interactive exploration, Python scripts for utility functions, and deep learning frameworks (likely PyTorch as indicated in chapter 2) utilized for the CNN models.