Getting Started¶
Welcome to Molfun. These guides walk you through everything you need to go from zero to fine-tuning protein structure models --- whether you are a computational biologist running your first prediction or an ML engineer integrating Molfun into a production pipeline.
No prior experience with AlphaFold or OpenFold required
The guides assume familiarity with Python and basic machine learning concepts. Knowledge of protein structure prediction is helpful but not necessary --- we explain the key ideas as we go.
Learning Path¶
Follow the guides in order for the smoothest experience, or jump to whichever step matches where you are.
1. Installation¶
Set up Python, install Molfun and its optional extras, and verify everything works.
2. Quick Start¶
Three code snippets to see what Molfun can do: structure prediction, LoRA fine-tuning, and the CLI --- all in under five minutes.
3. First Prediction¶
A step-by-step tutorial that loads a pretrained model, predicts a 3D structure, explores the output tensors, saves a PDB file, and visualizes the result.
4. First Fine-Tuning¶
End-to-end LoRA fine-tuning: prepare a dataset, pick a strategy, run model.fit(), track
the experiment, and evaluate the result.
What Comes Next?¶
Once you have completed these guides you will be able to:
- Predict protein structures, properties, and binding affinities
- Fine-tune models with LoRA, head-only, partial, or full strategies
- Save, load, and export models for production
From here, explore the Tutorials for real-world workflows, the Architecture docs to understand how the pieces fit together, or the API Reference for complete method signatures.