Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SeedProteo: Accurate De Novo All-Atom Design of Protein Binders

Wei Qu, Yiming Ma, Fei Ye, Chan Lu, Yi Zhou, Kexin Zhang, Lan Wang, Minrui Gui, Quanquan Gu

TL;DR

This work demonstrates how to repurpose a cutting-edge folding architecture into a powerful generative design framework by effectively integrating self-conditioning features, and validates SeedProteo through wet-lab assays on two therapeutic targets, establishing leading results.

Abstract

We present SeedProteo, a diffusion-based model for de novo all-atom protein design. We demonstrate how to repurpose a cutting-edge folding architecture into a powerful generative design framework by effectively integrating self-conditioning features. Extensive benchmarks highlight the model's capabilities across two distinct tasks: in unconditional generation, SeedProteo exhibits superior length generalization and structural diversity, maintaining robustness for long sequences and complex topologies; in binder design, it achieves state-of-the-art performance among open-source methods, attaining the highest in-silico design success rates, structural diversity and novelty. Finally, we validate SeedProteo through wet-lab assays on two therapeutic targets, achieving hit rates of 70%-80% and picomolar-level binding affinities, establishing leading results. To facilitate community adoption, we provide public access to SeedProteo via a webserver (https://seedfold.io/proteinDesign).

SeedProteo: Accurate De Novo All-Atom Design of Protein Binders

TL;DR

This work demonstrates how to repurpose a cutting-edge folding architecture into a powerful generative design framework by effectively integrating self-conditioning features, and validates SeedProteo through wet-lab assays on two therapeutic targets, establishing leading results.

Abstract

We present SeedProteo, a diffusion-based model for de novo all-atom protein design. We demonstrate how to repurpose a cutting-edge folding architecture into a powerful generative design framework by effectively integrating self-conditioning features. Extensive benchmarks highlight the model's capabilities across two distinct tasks: in unconditional generation, SeedProteo exhibits superior length generalization and structural diversity, maintaining robustness for long sequences and complex topologies; in binder design, it achieves state-of-the-art performance among open-source methods, attaining the highest in-silico design success rates, structural diversity and novelty. Finally, we validate SeedProteo through wet-lab assays on two therapeutic targets, achieving hit rates of 70%-80% and picomolar-level binding affinities, establishing leading results. To facilitate community adoption, we provide public access to SeedProteo via a webserver (https://seedfold.io/proteinDesign).
Paper Structure (39 sections, 4 figures, 6 tables)

This paper contains 39 sections, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Model Architecture
  4. Self-conditioning features
  5. Training Objective
  6. Training Pipeline
  7. In-silico Results
  8. Unconditional Generation Benchmark
  9. Scalability and Structural Diversity
  10. Fine-grained Evaluation across Structural Topologies
  11. Binder Design Benchmark
  12. Performance Analysis.
  13. Discussion on Co-Design Strategies
  14. Beyond the Inverse Folding Bottleneck
  15. Decoding Mechanisms
  16. ...and 24 more sections

Figures (4)

  • Figure 1: Overview of the SeedProteo framework. The right panel illustrates the input representation from a folding perspective. By modifying specific input channels within a nearly identical network architecture, we adapt the framework into the generative design model shown on the left.
  • Figure 2: Unconditional monomer benchmark. Stricter thresholds are applied to define designability (in Appendix \ref{['app:eval_settings']}).
  • Figure 3: Binders generated by SeedProteo for the challenging multi-chain targets H1 (dimer), VEGF-A (dimer), and TNF-$\alpha$ (trimer). All displayed binders, colored in purple, meet the in silico success criteria.
  • Figure 4: Wet-lab validation of SeedProteo-designed binders. For each target (PD-L1, top; SC2RBD, bottom), three representative binding complex structures are shown alongside the SPR sensorgram of the highest-affinity binder.