How se7en69/proteinscape works
Project type and purpose
• ProteinScape is a lightweight, web‐based full-stack application built with Streamlit.
File structure
• Top-level files:
– app.py: contains the single Streamlit application (both front-end and glue code for backend calls).
– requirements.txt: lists all dependencies (Streamlit, py3Dmol, Matplotlib, Seaborn, Pandas, Requests, NumPy).
– README.md / LICENSE: documentation and license.
• No separate “frontend” or “backend” directories: the architecture is a monolithic Streamlit service that integrates UI, data handling, plotting, and external API calls.
README for additional insights
• Frontend: Streamlit app.py
• Backend integration: Python requests to ESMFold API
• Visualization: py3Dmol for 3D, Matplotlib/Seaborn for 2D plots
• Data handling: Pandas, built-in collections
• File I/O: FASTA upload, PDB download
A. Client (Web Browser)
B. Streamlit Service (app.py)
1. UI layer (form inputs, buttons)
2. Controller logic (handles button clicks, file uploads)
3. API client (Requests → ESMFold API)
4. Data processor (Pandas, Collections)
5. Plot generator (Matplotlib, Seaborn)
6. 3D renderer (py3Dmol)
C. External Service
– ESMFold API (remote protein structure predictor)
D. File System / User Assets
– Uploaded FASTA files
– Downloadable PDB files
– Browser → Streamlit Service (HTTP/WebSocket calls handled by Streamlit)
– Streamlit Service → ESMFold API (HTTPS REST request)
– ESMFold API → Streamlit Service (PDB response)
– Streamlit Service → Browser (3D viewer, plots, downloads)
Architectural patterns and design principles
• Single-page interactive web application
• Monolithic service with clearly separated internal modules (UI vs. API vs. data vs. viz)
• External API integration for compute-intensive work (protein folding)
• Client-server separation: all business logic lives on the server (Streamlit), browser only renders results