How yasirbarlas/protein-secondary-structure-prediction works
Project Identification
• Type & Purpose: This is an academic/data-science project aimed at comparing three machine learning approaches—k-nearest neighbours, random forests, and neural networks—for predicting the secondary structure of proteins. The project uses notebooks for exploratory data analysis (EDA) and to experiment with various ML algorithms.
• Indicators: The README clearly describes the domain (bioinformatics & protein structure prediction) and notes that the repository supports coursework. The file tree is organized around experiments with different ML methods.
Analyzing the File Structure
• Top-Level Components:
• README.md: Provides the project’s overview, background, and methodology.
• background: Contains an EDA notebook to explore and preprocess data.
• datasets: Holds the raw and pre-processed CSV files with protein sequences.
• k-nearest-neighbours: Contains notebooks for building k-NN models.
• random-forests: Contains notebooks for training various random forest configurations.
• neural-networks: Contains notebooks for neural network models (with oversampling and without oversampling).
• Architectural Patterns: The project uses the "notebook-driven" approach often seen in academic and experimental data science projects. It’s structured into directories that separate experimental pipelines for each ML method. There’s clear separation of concerns: data storage (datasets), data preprocessing (background), and modeling experiments (separate folders per algorithm).
Putting It All Together
• “Data Sources” to an “EDA & Preprocessing” block (background/eda-data-processing.ipynb).
• Surround the ML modeling components with annotations that indicate the use of specific technology stacks (numpy, scikit-learn, TensorFlow) and any hardware specifics (e.g., GPU for neural nets).