So happy to share! We designed from scratch barrels with 8 beta-strands that fold in lipid membranes. To succeed, we had to extensively use negative design and destabilize the beta-strands to maintain the proteins unfolded in water. science.sciencemag.org/content/371/65…

A #deeplearning approach to protein design: When combined with detailed physical models, AI can capture properties normally only accessible through large-scale simulations. Protein sequence design by conformational landscape optimization pnas.org/content/118/11… | PNASNews

Almost there! 😎 Some similarity to Alphafold2, with an interesting twist: The 3D structure is generated in parallel with 1D (sequence), 2D (contacts), 3D (coordinates), all parts talking to each other iteratively.

OK, enough background. Over 600 rounds of selection (~3,000 generations), snowflake yeast evolved to become macroscopic. We’re talking mm scale- bigger than a fruit fly. They go from ~100 cells per cluster (left tube) to nearly half a million cells per cluster (right tube)! 12/35

If I could snap my fingers and change everything, here is what I would do: 1) Have all publishing happen on a single, consolidated, 100% subsidized "preprint" platform that publishes, formats, disseminates and archives all works of science with zero cost to authors or users.

1/5 Robust deep learning based protein sequence design using ProteinMPNN biorxiv.org/content/10.110… The high rate of experimental success and applicability to almost any protein sequence design problem has made it the standard approach at the Institute for Protein Design github.com/dauparas/Prote…

1/3 Excited to share our work on designing new protein assemblies w/ #AlphaFold hallucination and #ProteinMPNN. We validated 10 structures w/ up to 1550 residues and C33 symmetry. Work w/ @Milles_Lukas A. Courbet @RobertRagotte Justas Dauparas et al. biorxiv.org/content/10.110…

We’re hiring Lab Techs/Research Assistants in Copenhagen to help transform the latest advances in protein design & high-throughput cell/mol biology into therapies. Looking to expand our DK/USA team at BioInnovation Institute for a future startup! linkedin.com/jobs/search/?f…

ProteinMPNN paper: science.org/doi/10.1126/sc… Github: github.com/dauparas/Prote… Colab notebook: github.com/dauparas/Prote… HuggingFace notebook: huggingface.co/spaces/simondu… Hallucinating symmetric protein assemblies paper: science.org/doi/10.1126/sc… Thanks to all collaborators!

Excited to share our work on "top-down" reinforcement learning based approach to design proteins with exquisite structural control. The Monte Carlo tree search algorithm generate protein monomers and assemblies, guided by input geometric constraints. 1/4 biorxiv.org/content/10.110…

Lights, camera, enzymes! We designed new light-emitting luciferases using a DL method that generates fit-for-function proteins.

“I have not memorized all the protein structures,” says @uwproteindesign researcher Christoffer Norn. “The algorithms have.” The The New York Times looks at positive impacts of #AI, including protein design. Read: bit.ly/3MYlY5s

Designing proteins, next level, powered by #AI --Novel proteins by specification cell.com/chem/fulltext/… --Top-down science.org/doi/10.1126/sc… Science Magazine Institute for Protein Design Shunzhi Wang Christoffer Norn ISCRM at UW