Backbones

Sep 25, 2026

Surface-complementary seeds expand the structural diversity of diffusion-designed protein binders

Sep 07, 2026

Hard constraints can be enforced during diffusion by applying to fully denoised predictions and renoising

Aug 25, 2026

Protein backbone design methods show fold-level novelty by recycling memorized subdomain structures

Aug 19, 2026

Training inverse folding and diffusion models exclusively on predicted protein structures worsens performance due to how locally perfect they are

Jun 12, 2026

Joint sequence-structure diffusion or flow matching models show superior performance when designing structure at the beginning and sequence near the end

Apr 21, 2026

Inference-time scaling of de novo designed proteins is more effective for harder targets

Apr 21, 2026

AlphaFold2 outperforms ESMFold at distinguishing designable and undesignable protein backbones

Apr 21, 2026

Backbone diffusion works best on small protein sizes

Apr 21, 2026

Coupling sidechain and main chain prediction or design does not always lead to improvements

Apr 21, 2026

De novo designed backbones can increase the size of proteins for cryo-EM structure determination

Apr 21, 2026

De novo designed proteins have multiphasic folding pathways

Apr 21, 2026

De novo designed proteins with alpha helices are easier to predict than those with other secondary structures

Apr 21, 2026

Diffusion outperforms hallucination when using AF3-generation protein structure prediction methods

Apr 21, 2026

Diffusion-based protein design methods undersample structural diversity in specific topologies

Apr 21, 2026

Hallucination outperforms diffusion on protein refolding accuracy, particularly among larger proteins

Apr 21, 2026

Low-pLDDT hallucinated models from AF3-generation protein structure predictors are designable

Apr 21, 2026

Most possible alpha-beta topologies unexplored by nature

Apr 21, 2026

Multimodal sequence-structure guidance outperforms unimodal guidance in cases where both need to be designed

Apr 21, 2026

Nearly half of CATH domains do not pass designability filters used to evaluate protein backbone design performance

Apr 21, 2026

No one-size-fits-all best approach to motif scaffolding protein design

Apr 21, 2026

Partial diffusion of protein structures reduces structural complexity

design/backbones

Apr 21, 2026

Partial structure diffusion can make de novo backbones more designable

Apr 21, 2026

Protein backbone diffusion models undersample loop-rich and alpha-beta domains and functional motifs

Apr 21, 2026

Protein backbones designed by diffusion, but not by language models, have more secondary structure

Apr 21, 2026

Protein backbones designed using diffusion, but not sequence-based models, have fewer beta sheets

Apr 21, 2026

Proteins designed by diffusion are more compact than those designed by hallucination

design/backbones

Apr 21, 2026

Sheets are less designable than helices

Apr 21, 2026

Subquadratic scaling for protein backbone diffusion

Apr 21, 2026

Training backbone diffusion models on synthetic data improves designability

Apr 21, 2026

Text-to-structure protein design models outperform text-to-sequence models

Apr 21, 2026

Buried nonpolar surface area is a major determinant of whether de novo designed proteins are stable

Apr 21, 2026

Mutation of allosteric residues can influence the equilibrium dynamics of designed proteins

Apr 20, 2026

Protein backbone design

design/backbones

Quartz 4

Explorer

Backbones

Surface-complementary seeds expand the structural diversity of diffusion-designed protein binders

Hard constraints can be enforced during diffusion by applying to fully denoised predictions and renoising

Protein backbone design methods show fold-level novelty by recycling memorized subdomain structures

Training inverse folding and diffusion models exclusively on predicted protein structures worsens performance due to how locally perfect they are

Joint sequence-structure diffusion or flow matching models show superior performance when designing structure at the beginning and sequence near the end

Inference-time scaling of de novo designed proteins is more effective for harder targets

AlphaFold2 outperforms ESMFold at distinguishing designable and undesignable protein backbones

Backbone diffusion works best on small protein sizes

Coupling sidechain and main chain prediction or design does not always lead to improvements

De novo designed backbones can increase the size of proteins for cryo-EM structure determination

De novo designed proteins have multiphasic folding pathways

De novo designed proteins with alpha helices are easier to predict than those with other secondary structures

Diffusion outperforms hallucination when using AF3-generation protein structure prediction methods

Diffusion-based protein design methods undersample structural diversity in specific topologies

Hallucination outperforms diffusion on protein refolding accuracy, particularly among larger proteins

Low-pLDDT hallucinated models from AF3-generation protein structure predictors are designable

Most possible alpha-beta topologies unexplored by nature

Multimodal sequence-structure guidance outperforms unimodal guidance in cases where both need to be designed

Nearly half of CATH domains do not pass designability filters used to evaluate protein backbone design performance

No one-size-fits-all best approach to motif scaffolding protein design

Partial diffusion of protein structures reduces structural complexity

Partial structure diffusion can make de novo backbones more designable

Protein backbone diffusion models undersample loop-rich and alpha-beta domains and functional motifs

Protein backbones designed by diffusion, but not by language models, have more secondary structure

Protein backbones designed using diffusion, but not sequence-based models, have fewer beta sheets

Proteins designed by diffusion are more compact than those designed by hallucination

Sheets are less designable than helices

Subquadratic scaling for protein backbone diffusion

Training backbone diffusion models on synthetic data improves designability

Text-to-structure protein design models outperform text-to-sequence models

Buried nonpolar surface area is a major determinant of whether de novo designed proteins are stable

Mutation of allosteric residues can influence the equilibrium dynamics of designed proteins

Protein backbone design