FrameDiff

FrameDiff SE(3) diffusion

ICML 2023 main track (A*)

Summary card

ПолеЗначение
Год / version2023 / NR
Last verified2026-08-13
Что генерируетProtein backbone
ФорматыMonomer backbones (reported up to 500 aa)
De novo или optimizationDe novo
DOINR
Paperhttps://proceedings.mlr.press/v202/yim23a.html
GitHub / repohttps://github.com/jasonkyuyim/se3_diffusion
Service / docsNR
ЛицензияMIT (repository); bundled dependency terms apply
Доступность / commercial-useCode

Задача и generated object

Used in these tasks: De novo генерация backbone

Part of these workflows: De novo backbone → inverse folding

ПолеЗначение
Required inputLength specification
ConditioningPrimarily length; research implementation
FixedNR
EditableNR
GeneratedProtein backbone
Single mutationsNo
Combinatorial mutationsSequence must be added externally
Multi-objectiveNo

Architecture and operating principle

Author-reported (matrix)
ПолеЗначение
Base architectureNR
Model type3 Backbone/structure generator
RepresentationNR
Generation unitProtein backbone
ConditioningPrimarily length; research implementation
OutputProtein backbone
SamplingNR
Internal scorerDesignability metrics only
Structure generationNo

Input representation

NR

ПолеЗначение
Node featuresNR
Edge / pair featuresNR

Encoder и decoder

NR

ПолеЗначение
EncoderNR
DecoderNR

Latent or generative state

NR

Conditioning mechanism

Primarily length; research implementation

Generation mechanism

SE(3) diffusion / score model over residue frames

Sampling: NR

Training objectives

NR

Losses: NR

Mathematical details: NR

Integrated evaluation

Designability metrics only

Architectural limitations

Uses these architecture components: Frame diffusion, SE(3)-equivariant networks

Generator versus evaluation

ComponentRoleArchitecture classIntegrated or externalRequired or optionalInfluences generation or post hocRuntime included or excluded
FrameDiffprimary_generatorSE(3) diffusion / score model over residue framesintegratedrequiredgenerationincluded
Designability metrics onlyintegrated_scorerNRintegratedoptionalpost_hocunknown
ProteinMPNN + structure prediction/self-consistencymandatory_external_dependencyNRexternalrequiredpost_hocexcluded
Experimental binding/function assaysexperimental_validationNRexternalrequiredpost_hocexcluded

Training data и biases

ПолеЗначение
Training dataNR
PreprocessingNR
Negative examplesNR
Fine-tuningNR
Label requirementsNR
Training hardwareNR
Training timeNR

Inference hardware и runtime

ПолеЗначение
Candidate countNR
Batch sizeNR
Generation stepsNR
RecyclesNR
SizeNR
Reported runtimeNR
Runtime boundaryNR
CPUNR
GPUNR
VRAMNR
RAMNR
DiskNR
DatabasesNR

Experimental evidence

ПолеЗначение
Evidence class1 in silico generation/evaluation
RetrospectiveNo prospective wet-lab validation in primary paper
Prospective bindingNR
Prospective functionNR
DevelopabilityNR
Tested design countNo prospective wet-lab validation in primary paper
Success criterionComputational designability and novelty
Independent validationNR
SummaryComputational designability and novelty

Limitations and failure modes

Deployment: Research-only

Reproducibility, license, commercial use

ПолеЗначение
Repohttps://github.com/jasonkyuyim/se3_diffusion
LicenseMIT (repository); bundled dependency terms apply
Commercial-use statusCode

Alternative tools: NR

Complementary: NR

Related: NR

Not directly comparable with

AntiFold, Efficient evolution with ESM-1v, ESM-IF1, EvoDiff, LigandMPNN, ProGen / ProGen2, ProteinMPNN, Structural evolution (ESM-IF1 complex)
Сравнение допустимо только внутри совместимых категорий. Runtime и quality AntiFold vs RFantibody нельзя читать как одну шкалу: разные starting points и generated objects.

Depends on these external modules

NR

NR

Primary paper и полный список источников

SE(3) diffusion model with application to protein backbone generation

Source matrix: landscape xlsx, last verified 2026-08-13.

Детальный source-grounded разбор

sources identified Последняя сверка источников: 2026-08-27. Ниже — каркас обязательных разделов, заполненный из канонического YAML инструмента. Это не замена чтения primary paper / SI / model files. Поля без первоисточника помечены NR или MATRIX.

1. Что именно делает инструмент

[MATRIX]Вход: Length specification

[MATRIX]Generated object: Protein backbone

[MATRIX]Фиксируется: NR. Изменяется: NR.

[MATRIX]De novo или parent-based: De novo.

2. Архитектура в одном абзаце

[MATRIX]SE(3) diffusion / score model over residue frames

4. Математическая модель

NRТочные уравнения, losses и sampling solver будут внесены после чтения paper/code. Не подставляйте родовой UCB/transformer/diffusion, если он не подтверждён для этого инструмента.

Поля NR: conditional probability, update equations, sampling algorithm

5. Физическое устройство состояния

[MATRIX]Seed generated state / output: Protein backbone. Structure generation: No.

6. Физико-химическая интерпретация

[INTERPRETATION]Координаты, frames, PLM embeddings или residue tokens не являются расчётом ΔG, электростатики или растворителя, пока это не сделано внешней energy function. Likelihood / pLDDT / docking score ≠ экспериментальный binding.

7. Training

[MATRIX]Данные: NR. Fine-tuning: NR. Negative examples: NR.

8. Inference

[MATRIX]Sampling: NR. Generation steps: NR. Runtime boundary: NR.

10. Почему, по мнению авторов, метод эффективен

[MATRIX]Author-reported summary (матрица): Computational designability and novelty

NRCausal support не классифицирован, пока не прочитаны ablation и experimental sections.

12. Ablation analysis

NRAblation в этом выпуске не разобрана.

13. What the model does not represent

14. Evidence and limitations

[MATRIX]Evidence class (матрица): 1 in silico generation/evaluation. Prospective binding: NR. Function: NR. Developability: NR.

15. Reproducibility

NRРепозиторий: https://github.com/jasonkyuyim/se3_diffusion. Лицензия: MIT (repository); bundled dependency terms apply. Commit: NR, пока страница не имеет отдельного analysis YAML.

Модульная схема

МодульКласс моделиВходВыходОбучается или frozenGenerator или evaluator
FrameDiffSE(3) diffusion / score model over residue framesNRNRNRprimary_generator
Designability metrics onlyNRNRNRNRintegrated_scorer
ProteinMPNN + structure prediction/self-consistencyNRNRNRNRmandatory_external_dependency
Experimental binding/function assaysNRNRNRNRexperimental_validation

Generator versus evaluator

КомпонентРольIntegrated или externalRequired или optionalВлияет на generation или только ранжирует после
FrameDiffprimary_generatorintegratedrequiredgeneration
Designability metrics onlyintegrated_scorerintegratedoptionalpost_hoc
ProteinMPNN + structure prediction/self-consistencymandatory_external_dependencyexternalrequiredpost_hoc
Experimental binding/function assaysexperimental_validationexternalrequiredpost_hoc

Claim–evidence matrix

УтверждениеПредполагаемый механизмТип evidenceДатасет / экспериментComparatorМетрика / результатИсточникОговорка
Computational designability and noveltySE(3) diffusion / score model over residue frames2. retrospective benchmarkNo prospective wet-lab validation in primary paperNRComputational designability and novelty[MATRIX] landscape matrix 2026-08-13Не читать как полный разбор primary paper.