EvoDiff

Preprint

Summary card

Поле	Значение
Год / version	2023/2024 / NR
Last verified	2026-08-13
Что генерирует	Amino-acid sequence
Форматы	Protein sequences, including IDRs
De novo или optimization	De novo or masked local redesign
DOI	10.1101/2023.09.11.556673
Paper	https://doi.org/10.1101/2023.09.11.556673
GitHub / repo	https://github.com/microsoft/evodiff
Service / docs	NR
Лицензия	MIT
Доступность / commercial-use	Code + weights

Задача и generated object

Used in these tasks: Генерация белковых sequences, Motif scaffolding

Part of these workflows: Sequence-first de novo generation

Поле	Значение
Required input	None, partial sequence/motif, or MSA
Conditioning	Sequence motif masks; MSA context; supports disordered regions
Fixed	NR
Editable	NR
Generated	Amino-acid sequence
Single mutations	Yes via masks
Combinatorial mutations	Yes
Multi-objective	No explicit objective optimizer

Architecture and operating principle

Author-reported (matrix)

Поле	Значение
Base architecture	NR
Model type	1 Sequence generator
Representation	NR
Generation unit	Amino-acid sequence
Conditioning	Sequence motif masks; MSA context; supports disordered regions
Output	Amino-acid sequence
Sampling	NR
Internal scorer	No task-specific internal ranker
Structure generation	No

Input representation

Поле	Значение
Node features	NR
Edge / pair features	NR

Encoder и decoder

Поле	Значение
Encoder	NR
Decoder	NR

Latent or generative state

Conditioning mechanism

Sequence motif masks; MSA context; supports disordered regions

Generation mechanism

Discrete diffusion over amino-acid tokens; OA-DM / D3PM variants

Sampling: NR

Training objectives

Losses: NR

Mathematical details: NR

Integrated evaluation

No task-specific internal ranker

Architectural limitations

Sequence-only generator.
motif presence does not establish correct 3D placement or function.
prospective evidence absent.

Uses these architecture components: Sequence diffusion, Masked-token infilling

Generator versus evaluation

Component	Role	Architecture class	Integrated or external	Required or optional	Influences generation or post hoc	Runtime included or excluded
Discrete sequence diffusion	primary_generator	Discrete diffusion over amino-acid tokens; OA-DM / D3PM variants	integrated	required	generation	included
No task-specific internal ranker	integrated_scorer	NR	integrated	optional	post_hoc	unknown
Structure prediction and assay-specific validation	mandatory_external_dependency	NR	external	required	post_hoc	excluded
Experimental binding/function assays	experimental_validation	NR	external	required	post_hoc	excluded

Training data и biases

Поле	Значение
Training data	NR
Preprocessing	NR
Negative examples	NR
Fine-tuning	NR
Label requirements	NR
Training hardware	NR
Training time	NR

Inference hardware и runtime

Поле	Значение
Candidate count	NR
Batch size	NR
Generation steps	NR
Recycles	NR
Size	NR
Reported runtime	NR
Runtime boundary	NR
CPU	NR
GPU	NR
VRAM	NR
RAM	NR
Disk	NR
Databases	NR

Experimental evidence

Поле	Значение
Evidence class	1–2 in silico / retrospective
Retrospective	No prospective wet-lab validation in primary work
Prospective binding	NR
Prospective function	NR
Developability	NR
Tested design count	No prospective wet-lab validation in primary work
Success criterion	Computational sequence, structural-plausibility and motif-scaffolding evaluation
Independent validation	NR
Summary	Computational sequence, structural-plausibility and motif-scaffolding evaluation

Limitations and failure modes

Sequence-only generator.
motif presence does not establish correct 3D placement or function.
prospective evidence absent.

Deployment: Research-only

Reproducibility, license, commercial use

Поле	Значение
Repo	https://github.com/microsoft/evodiff
License	MIT
Commercial-use status	Code + weights

Alternative tools: NR

Complementary: NR

Related: NR

Not directly comparable with

AntiFold, ESM-IF1, FrameDiff, LigandMPNN, ProteinMPNN, RFdiffusion, RFdiffusion All-Atom (RFdiffusionAA)

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

Primary paper и полный список источников

Protein generation with evolutionary diffusion: sequence is all you need

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

Детальный source-grounded разбор

1. Что именно делает инструмент

[MATRIX]Вход: None, partial sequence/motif, or MSA

[MATRIX]Generated object: Amino-acid sequence

[MATRIX]Фиксируется: NR. Изменяется: NR.

[MATRIX]De novo или parent-based: De novo or masked local redesign.

2. Архитектура в одном абзаце

[MATRIX]Discrete diffusion over amino-acid tokens; OA-DM / D3PM variants

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: Amino-acid sequence. 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 sequence, structural-plausibility and motif-scaffolding evaluation

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

Явный растворитель и ионная сила — обычно отсутствуют в генераторе.
Протонирование, гликозилирование, кофакторы — обычно отсутствуют, если не оговорено.
Конформационная энтропия и долгая динамика — не моделируются одним forward pass.
Экспрессия, агрегация, иммуногенность — не равны likelihood генератора.
Истинная binding free energy — не равна internal score.

14. Evidence and limitations

[MATRIX]Evidence class (матрица): 1–2 in silico / retrospective. Prospective binding: NR. Function: NR. Developability: NR.

Sequence-only generator.
motif presence does not establish correct 3D placement or function.
prospective evidence absent.

15. Reproducibility

NRРепозиторий: https://github.com/microsoft/evodiff. Лицензия: MIT. Commit: NR, пока страница не имеет отдельного analysis YAML.

Модульная схема

Модуль	Класс модели	Вход	Выход	Обучается или frozen	Generator или evaluator
Discrete sequence diffusion	Discrete diffusion over amino-acid tokens; OA-DM / D3PM variants	NR	NR	NR	primary_generator
No task-specific internal ranker	NR	NR	NR	NR	integrated_scorer
Structure prediction and assay-specific validation	NR	NR	NR	NR	mandatory_external_dependency
Experimental binding/function assays	NR	NR	NR	NR	experimental_validation

Generator versus evaluator

Компонент	Роль	Integrated или external	Required или optional	Влияет на generation или только ранжирует после
Discrete sequence diffusion	primary_generator	integrated	required	generation
No task-specific internal ranker	integrated_scorer	integrated	optional	post_hoc
Structure prediction and assay-specific validation	mandatory_external_dependency	external	required	post_hoc
Experimental binding/function assays	experimental_validation	external	required	post_hoc

Claim–evidence matrix

Утверждение	Предполагаемый механизм	Тип evidence	Датасет / эксперимент	Comparator	Метрика / результат	Источник	Оговорка
Computational sequence, structural-plausibility and motif-scaffolding evaluation	Discrete diffusion over amino-acid tokens; OA-DM / D3PM variants	2. retrospective benchmark	No prospective wet-lab validation in primary work	NR	Computational sequence, structural-plausibility and motif-scaffolding evaluation	[MATRIX] landscape matrix 2026-08-13	Не читать как полный разбор primary paper.