A field guide to engineering proteins with neural nets

What it does

This repository is a curated index of academic papers applying deep learning to protein design. The maintainer collects research spanning benchmarks, reviews, and concrete design tasks—such as de novo proteins, antibodies, enzymes, and binders—and sorts them into a taxonomy of model architectures and biological workflows. It functions as a map of how neural networks are being repurposed to engineer biological molecules.

The interesting bit

Rather than a flat list, the repository groups papers by methodological approach—diffusion-based, LLM-based, GAN-based, flow-based, reinforcement learning, and more—making it easy to trace which architectural trends are colonizing which biological problems. It also surfaces associated code repositories and datasets when available, turning a literature review into a launchpad.

Key highlights

• Covers the full design pipeline: function-to-scaffold, scaffold-to-sequence, function-to-sequence, and function-to-structure.
• Maintains a weekly additions section linking to the latest preprints and publications.
• Includes benchmarks and datasets such as FLIP, ProteinGym, and PDBench with links to their code.
• Maintains a separate section for responsible AI development commitments in protein design.

Verdict

Bookmark this if you work where machine learning meets structural biology and need to survey methods fast; avoid it if you want runnable code or tutorials—this is strictly a reading list.