BoltzGen

Preprint Prospective flag

Summary card

Поле	Значение
Год / version	2025 / NR
Last verified	2026-08-13
Что генерирует	Binder sequence–structure and target complex
Форматы	Protein, peptide, disulfide peptide, nanobody; complexes
De novo или optimization	De novo
DOI	10.1101/2025.11.20.689494
Paper	https://doi.org/10.1101/2025.11.20.689494
GitHub / repo	https://github.com/HannesStark/boltzgen
Service / docs	NR
Лицензия	MIT
Доступность / commercial-use	Training/inference code + weights/designs

Задача и generated object

Used in these tasks: Target-conditioned binder generation, De novo VHH / nanobody design, Joint sequence–structure generation

Part of these workflows: All-atom universal binder generation (BoltzGen)

Поле	Значение
Required input	Target biomolecule plus design specification
Conditioning	Residue/chain masks, modality, target context, geometry; supports diverse biomolecules
Fixed	NR
Editable	NR
Generated	Binder sequence–structure and target complex
Single mutations	Not primary
Combinatorial mutations	Full sequence
Multi-objective	Specification/filter-based; explicit Pareto support NR

Architecture and operating principle

Author-reported (matrix)

Поле	Значение
Base architecture	NR
Model type	4 + 5 Joint sequence–structure and target-conditioned binder generator
Representation	NR
Generation unit	Binder sequence–structure and target complex
Conditioning	Residue/chain masks, modality, target context, geometry; supports diverse biomolecules
Output	Binder sequence–structure and target complex
Sampling	NR
Internal scorer	Integrated folding/ranking; affinity components depend on protocol
Structure generation	Yes—generation and structure prediction unified

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

Residue/chain masks, modality, target context, geometry; supports diverse biomolecules

Generation mechanism

All-atom generative diffusion unified with folding; design specification language

Sampling: NR

Training objectives

Losses: NR

Mathematical details: NR

Integrated evaluation

Integrated folding/ranking; affinity components depend on protocol

Architectural limitations

Preprint.
unified model scores are not independent validation.
evidence differs greatly by binder modality.

Uses these architecture components: Joint sequence–structure diffusion, Coordinate diffusion, 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
BoltzGen all-atom diffusion	primary_generator	All-atom generative diffusion unified with folding; design specification language	integrated	required	generation	included
Integrated folding/ranking; affinity components depend on protocol	integrated_scorer	NR	integrated	optional	post_hoc	unknown
Yes—generation and structure prediction unified	integrated_structure_predictor	NR	integrated	optional	post_hoc	included
Independent prediction and experimental binding/function 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	3–4 prospective binding/functional validation
Retrospective	NR
Prospective binding	Includes nanobody and live-cell/functional examples; still preprint
Prospective function	Includes nanobody and live-cell/functional examples; still preprint
Developability	NR
Tested design count	Distributed multi-lab campaigns across modalities; per-task counts in preprint
Success criterion	Includes nanobody and live-cell/functional examples; still preprint
Independent validation	NR
Summary	Includes nanobody and live-cell/functional examples; still preprint

Limitations and failure modes

Preprint.
unified model scores are not independent validation.
evidence differs greatly by binder modality.

Deployment: Frontier; limited pilot after reproduction

Reproducibility, license, commercial use

Поле	Значение
Repo	https://github.com/HannesStark/boltzgen
License	MIT
Commercial-use status	Training/inference code + weights/designs

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

BoltzGen: universal binder design

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

Review queue

Bibliographic title must be verified against the PDF.

Детальный source-grounded разбор

1. Что именно делает инструмент

[MATRIX]Вход: Target biomolecule plus design specification

[MATRIX]Generated object: Binder sequence–structure and target complex

[MATRIX]Фиксируется: NR. Изменяется: NR.

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

2. Архитектура в одном абзаце

[MATRIX]All-atom generative diffusion unified with folding; design specification language

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 sequence–structure and target complex. Structure generation: Yes—generation and structure prediction unified.

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): Includes nanobody and live-cell/functional examples; still preprint

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: Includes nanobody and live-cell/functional examples; still preprint. Function: Includes nanobody and live-cell/functional examples; still preprint. Developability: NR.

Preprint.
unified model scores are not independent validation.
evidence differs greatly by binder modality.

15. Reproducibility

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

Модульная схема

Модуль	Класс модели	Вход	Выход	Обучается или frozen	Generator или evaluator
BoltzGen all-atom diffusion	All-atom generative diffusion unified with folding; design specification language	NR	NR	NR	primary_generator
Integrated folding/ranking; affinity components depend on protocol	NR	NR	NR	NR	integrated_scorer
Yes—generation and structure prediction unified	NR	NR	NR	NR	integrated_structure_predictor
Independent prediction and experimental binding/function 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
BoltzGen all-atom diffusion	primary_generator	integrated	required	generation
Integrated folding/ranking; affinity components depend on protocol	integrated_scorer	integrated	optional	post_hoc
Yes—generation and structure prediction unified	integrated_structure_predictor	integrated	optional	post_hoc
Independent prediction and experimental binding/function 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
Includes nanobody and live-cell/functional examples; still preprint	All-atom generative diffusion unified with folding; design specification language	5. prospective functional validation	Distributed multi-lab campaigns across modalities; per-task counts in preprint	NR	Includes nanobody and live-cell/functional examples; still preprint	[MATRIX] landscape matrix 2026-08-13	Не читать как полный разбор primary paper.