Drug Discovery Generative AI End-to-End De Novo Proteins; Tensor Networks Event Seminar and PDF 04/18/24.

Further modifications were made to a Ni, B., et al. Science Advances 2024 Property-to-sequence generation of de novo proteins using a language diffusion model notebook.
Property: Pulling forces learned from protein unfolding experiments using a prior language model.
Sequence: Generative AI diffusion model de novo arrangement of amino acids in the protein produced.
Authors’ previous property-to-sequence: Secondary structures specific to a protein are generated, Notebooks available.
End-to-end generative models aimed at creating proteins with a specific property are rare.
Authors include unfolding energy, mechanical strength with detailed unfolding force-separation curves.
Original pull force data for AI training comes from single-molecule technology such as atomic force microscopy, AFM.

A movie of the author's generated proteins being stretched is available through a literature Download.

“Here, we report a generative model that predicts protein designs to meet complex nonlinear mechanical property-design objectives. Our model leverages deep knowledge on protein sequences from a pretrained protein language model and maps mechanical unfolding responses to create proteins.”
- Ni, B., et al., Science Advances, February 07, 2024.