Hot Topics #32
Epitope-targeted de novo antibody design, Odyssey pLM, protein structure hallucination and more
Efficient generation of epitope-targeted de novo antibodies with Germinal: Mille-Fragoso et al.: Sept. 24, 2025
Abstract: Obtaining novel antibodies against specific protein targets is a widely important yet experimentally laborious process. Meanwhile, computational methods for antibody design have been limited by low success rates that currently require resource-intensive screening. Here, we introduce Germinal, a broadly enabling generative framework that designs antibodies against specific epitopes with nanomolar binding affinities while requiring only low-throughput experimental testing. Our method co-optimizes antibody structure and sequence by integrating a structure predictor with an antibody-specific protein language model to perform de novo design of functional complementarity-determining regions (CDRs) onto a user-specified structural framework. When tested against four diverse protein targets, Germinal achieved an experimental success rate of 4-22% across all targets, testing only 43-101 designs for each antigen. Validated nanobodies also exhibited robust expression in mammalian cells and nanomolar binding affinities. We provide open-source code and full computational and experimental protocols to facilitate wide adoption. Germinal represents a milestone in low-throughput, epitope-targeted de novo antibody design, with notable implications for the development of molecular tools and therapeutics.
Odyssey: reconstructing evolution through emergent consensus in the global proteome: Singhal et al: October 15, 2025
Abstract: We present Odyssey, a family of multimodal protein language models for sequence and structure generation, protein editing and design. We scale Odyssey to more than 102 billion parameters, trained over 1.1 × 1023 FLOPs. The Odyssey architecture uses context modalities, categorized as structural cues, semantic descriptions, and orthologous group metadata, and comprises two main components: a finite scalar quantizer for tokenizing continuous atomic coordinates, and a transformer stack for multimodal representation learning. Odyssey is trained via discrete diffusion, and characterizes the generative process as a time-dependent unmasking procedure. The finite scalar quantizer and transformer stack leverage the consensus mechanism, a replacement for attention that uses an iterative propagation scheme informed by local agreements between residues. Across various benchmarks, Odyssey achieves landmark performance for protein generation and protein structure discretization. Our empirical findings are supported by theoretical analysis.
Stimuli-triggered formation of de novo-designed protein biomaterials: Gregorio et al.: Oct. 17, 2025
Abstract: Biomaterials are revolutionizing how we study biology and treat diseases, offering new platforms for tissue engineering, drug delivery, and cellular modulation. Among these, protein-based materials stand out for their ability to mimic biological environments with unmatched precision. Despite this, most protein hydrogels rely on a narrow set of naturally occurring building blocks, limiting their versatility. This work introduces a new frontier in biomaterials by leveraging de novo protein design, a computational approach that creates entirely new proteins from scratch. By engineering proteins that self-assemble into defined architectures and respond to external stimuli, such as small molecules, we demonstrate the creation of customizable bulk hydrogels and intracellular condensates with tunable mechanical properties and formation dynamics.
These materials not only expand the toolkit for bioengineers but also provide a powerful platform for probing fundamental biological processes. For example, the ability to trigger condensate formation inside cells opens new avenues for studying intracellular liquid-liquid phase separation, a phenomenon increasingly linked to aging and disease. Moreover, the modularity of this system, where protein components and triggers can be swapped, suggests broad applicability across biotechnology, synthetic biology, and regenerative medicine. As de novo design continues to evolve, it promises to unlock a vast landscape of protein-based materials with properties and functions beyond what nature has provided, reshaping how we build and interact with biological systems.
Multi-state Protein Design with DynamicMPNN: Abrudan et al.: July 29, 2025
Abstract: Structural biology has long been dominated by the one sequence, one structure, one function paradigm, yet many critical biological processes - from enzyme catalysis to membrane transport - depend on proteins that adopt multiple conformational states. Existing multi-state design approaches rely on post-hoc aggregation of single-state predictions, achieving poor experimental success rates compared to single-state design. We introduce DynamicMPNN, an inverse folding model explicitly trained to generate sequences compatible with multiple conformations through joint learning across conformational ensembles. Trained on 46,033 conformational pairs covering 75% of CATH superfamilies and evaluated using Alphafold 3, DynamicMPNN outperforms ProteinMPNN by up to 25% on decoy-normalized RMSD and by 12% on sequence recovery across our challenging multi-state protein benchmark.
Protein Hunter: exploiting structure hallucination within diffusion for protein design: Cho et al.: October 10, 2025
Abstract: Interactions between proteins and other biomolecules underlie nearly all biological processes, yet designing such interactions de novo remains challenging. Capturing their specific interactions and co-optimizing sequence and structure are difficult and often require extensive computation. We present Protein Hunter, a fast, fine-tuning-free framework for de novo protein design. Starting from an all-X sequence, we find diffusion-based structure prediction models hallucinate reasonable looking structures that can be further improved through iterative sequence re-design and structure re-prediction. This lightweight strategy achieves high AlphaFold3 in silico success rates across both unconditional and conditional generation tasks, including binders to proteins, cyclic peptides, small molecules, DNA, and RNA. Protein Hunter also supports multi-motif scaffolding and partial redesign, providing a general and efficient platform for de novo protein design across diverse molecular targets.
Proteinbase: “Today we’re launching Proteinbase, a single hub for experimental protein design data. Over 1,000 novel proteins are already live, each with computational predictions, experimental validation, and the method used to design them. Everything comes from the Adaptyv lab under standardized protocols, which means the results are reproducible, comparable, and include negative data that usually never gets shared.”
https://evedesign.bio/
