AlphaProteo

Preprint Technical report Prospective flag

Summary card

Поле	Значение
Год / version	2024 / NR
Last verified	2026-08-13
Что генерирует	Binder sequences/structures
Форматы	Small non-antibody protein binders
De novo или optimization	De novo
DOI	10.48550/arXiv.2409.08022
Paper	https://arxiv.org/abs/2409.08022
GitHub / repo	https://github.com/google-deepmind/ap_trusted_testers
Service / docs	NR
Лицензия	Trusted-tester notebooks Apache-2.0; model not released
Доступность / commercial-use	Restricted/trusted-tester; no local weights

Задача и generated object

Used in these tasks: De novo non-antibody binder design, Target-conditioned binder generation

Part of these workflows: Target-conditioned miniprotein binder (BindCraft)

Поле	Значение
Required input	Target structure / target region
Conditioning	Target surface/region; public controllability details limited
Fixed	NR
Editable	NR
Generated	Binder sequences/structures
Single mutations	No
Combinatorial mutations	Full sequence
Multi-objective	NR

Architecture and operating principle

Author-reported (matrix)

Поле	Значение
Base architecture	NR
Model type	5 Target-conditioned binder generator
Representation	NR
Generation unit	Binder sequences/structures
Conditioning	Target surface/region; public controllability details limited
Output	Binder sequences/structures
Sampling	NR
Internal scorer	Internal proprietary ranking
Structure generation	Undisclosed/closed

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

Target surface/region; public controllability details limited

Generation mechanism

Proprietary generative and ranking model family (technical report)

Sampling: NR

Training objectives

Losses: NR

Mathematical details: NR

Integrated evaluation

Internal proprietary ranking

Architectural limitations

Closed model and insufficient reproducibility.
not antibody-specific.
technical-report evidence.

Uses these architecture components: Joint sequence–structure generation

Generator versus evaluation

Component	Role	Architecture class	Integrated or external	Required or optional	Influences generation or post hoc	Runtime included or excluded
AlphaProteo (closed)	primary_generator	Proprietary generative and ranking model family (technical report)	integrated	required	generation	included
Internal proprietary ranking	integrated_scorer	NR	integrated	optional	post_hoc	unknown
Undisclosed/closed	integrated_structure_predictor	NR	integrated	optional	post_hoc	included
Experimental binding/functional assays still required	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	3–4 prospective binding/functional validation
Retrospective	NR
Prospective binding	Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark
Prospective function	Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark
Developability	NR
Tested design count	Seven targets in technical report
Success criterion	Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark
Independent validation	NR
Summary	Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark

Limitations and failure modes

Closed model and insufficient reproducibility.
not antibody-specific.
technical-report evidence.

Deployment: Frontier; not internally deployable

Reproducibility, license, commercial use

Поле	Значение
Repo	https://github.com/google-deepmind/ap_trusted_testers
License	Trusted-tester notebooks Apache-2.0; model not released
Commercial-use status	Restricted/trusted-tester; no local weights

Alternative tools: NR

Complementary: NR

Related: NR

Not directly comparable with

HuAbDiffusion, HuDiff, Humatch, RFantibody, tFold System

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

AlphaProteo generates novel proteins for biology and health

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

Детальный source-grounded разбор

1. Что именно делает инструмент

[MATRIX]Вход: Target structure / target region

[MATRIX]Generated object: Binder sequences/structures

[MATRIX]Фиксируется: NR. Изменяется: NR.

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

2. Архитектура в одном абзаце

[MATRIX]Proprietary generative and ranking model family (technical report)

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: Binder sequences/structures. Structure generation: Undisclosed/closed.

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): Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark

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): 3–4 prospective binding/functional validation. Prospective binding: Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark. Function: Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark. Developability: NR.

Closed model and insufficient reproducibility.
not antibody-specific.
technical-report evidence.

15. Reproducibility

NRРепозиторий: https://github.com/google-deepmind/ap_trusted_testers. Лицензия: Trusted-tester notebooks Apache-2.0; model not released. Commit: NR, пока страница не имеет отдельного analysis YAML.

Модульная схема

Модуль	Класс модели	Вход	Выход	Обучается или frozen	Generator или evaluator
AlphaProteo (closed)	Proprietary generative and ranking model family (technical report)	NR	NR	NR	primary_generator
Internal proprietary ranking	NR	NR	NR	NR	integrated_scorer
Undisclosed/closed	NR	NR	NR	NR	integrated_structure_predictor
Experimental binding/functional assays still required	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
AlphaProteo (closed)	primary_generator	integrated	required	generation
Internal proprietary ranking	integrated_scorer	integrated	optional	post_hoc
Undisclosed/closed	integrated_structure_predictor	integrated	optional	post_hoc
Experimental binding/functional assays still required	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
Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark	Proprietary generative and ranking model family (technical report)	5. prospective functional validation	Seven targets in technical report	NR	Reported 3–300× affinity improvements versus comparator methods; this is not a head-to-head universal benchmark	[MATRIX] landscape matrix 2026-08-13	Не читать как полный разбор primary paper.