A Generative Neuro‐Symbolic AI for Protein Sequence Design
Marianne Defresne, Delphine Dessaux, Samuel Buchet, Lucie Barthe, Liza Ammar‐Khodja, Bessam Azizi, Valentin Durante, Gianluca Cioci, Simon de Givry, Alain Roussel, Luis F Garcia‐Alles, Thomas Schiex, Sophie Barbe · 2026
Read A Generative Neuro‐Symbolic AI for Protein Sequence Design by Marianne Defresne, Delphine Dessaux, Samuel Buchet, Lucie Barthe, Liza Ammar‐Khodja, Bessam Azizi, Valentin Durante, Gianluca Cioci, Simon de Givry, Alain Roussel, Luis F Garcia‐Alles, Thomas Schiex, Sophie Barbe (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13422582.txt.
Description
Defresne and co-authors introduce a protein design system that marries deep learning with logical proof, in this 2026 study. The aim is to fix a known weakness of AI protein designers. They generate sequences residue by residue. They cannot think ahead about long-range interactions. Their framework, EffieDes, converts a neural network's sense of a target fold into a probabilistic model that a reasoning engine can search rigorously.
Validation came in two demanding tests. The system designed pairs of sequences that adopt the same fold. The pairs assemble only with their intended partner, a selectivity puzzle. It also produced a brand-new nanobody, a miniature antibody. The nanobody bound an immune-evasive SARS-CoV-2 variant with nanomolar affinity. The reasoning layer can enforce explicit constraints while optimizing the learned fitness score. So the authors argue the hybrid approach suits sophisticated functional objectives. Pure sampling fumbles those.
Catalog record
| CREATOR | Marianne Defresne, Delphine Dessaux, Samuel Buchet, Lucie Barthe, Liza Ammar‐Khodja, Bessam Azizi, Valentin Durante, Gianluca Cioci, Simon de Givry, Alain Roussel, Luis F Garcia‐Alles, Thomas Schiex, Sophie Barbe |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69 KB |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |