tFold System

tFold tFold-Ab tFold-Ag

Peer-reviewed Prospective flag Direct antibody flag

Summary card

Поле	Значение
Год / version	2025/2026 / NR
Last verified	2026-08-13
Что генерирует	CDR sequences, candidate antibody structures and antibody–antigen complexes
Форматы	Paired-chain mAb/Fv candidates; Fc-fusion/IgG-style experimental formats
De novo или optimization	De novo CDRs on fixed human framework
DOI	10.1038/s41467-025-67361-9
Paper	https://doi.org/10.1038/s41467-025-67361-9
GitHub / repo	https://github.com/TencentAI4S/tfold
Service / docs	NR
Лицензия	Custom Tencent license; commercial partnership requested
Доступность / commercial-use	Source available + weights/data; commercial review

Задача и generated object

Used in these tasks: Antigen-conditioned antibody design, Генерация CDR sequence, De novo antibody design

Part of these workflows: Epitope-specific mAb CDR generation (tFold)

Поле	Значение
Required input	Antigen sequence, specified epitope and human framework
Conditioning	Fixed framework family, epitope, CDR length combinations
Fixed	NR
Editable	NR
Generated	CDR sequences, candidate antibody structures and antibody–antigen complexes
Single mutations	No
Combinatorial mutations	Yes
Multi-objective	Filter cascade; no explicit Pareto optimizer

Architecture and operating principle

Author-reported (matrix)

Поле	Значение
Base architecture	NR
Model type	5 Target-conditioned antibody binder workflow
Representation	NR
Generation unit	CDR sequences, candidate antibody structures and antibody–antigen complexes
Conditioning	Fixed framework family, epitope, CDR length combinations
Output	CDR sequences, candidate antibody structures and antibody–antigen complexes
Sampling	NR
Internal scorer	tFold-Ag binding/complex prediction + virtual-screen filters
Structure generation	tFold-Ab and tFold-Ag integrated

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

Fixed framework family, epitope, CDR length combinations

Generation mechanism

Sequence generator + tFold-Ab/tFold-Ag structure/complex models + structure-aware filters

Sampling: NR

Training objectives

Losses: NR

Mathematical details: NR

Integrated evaluation

tFold-Ag binding/complex prediction + virtual-screen filters

Architectural limitations

Generation, ranking and display selection are intertwined.
reported affinities cannot be attributed to generator alone.
framework/design space constrained.

Uses these architecture components: Protein language models, Geometric transformers

Generator versus evaluation

Component	Role	Architecture class	Integrated or external	Required or optional	Influences generation or post hoc	Runtime included or excluded
Sequence-based antibody generator	primary_generator	Sequence generator + tFold-Ab/tFold-Ag structure/complex models + structure-aware filters	integrated	required	generation	included
tFold-Ag binding/complex prediction + virtual-screen filters	integrated_scorer	NR	integrated	optional	post_hoc	unknown
tFold-Ab and tFold-Ag integrated	integrated_structure_predictor	NR	integrated	optional	post_hoc	included
Independent complex validation; phage display/SPR/competition assays	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	3–5 prospective binding, epitope and developability evidence
Retrospective	NR
Prospective binding	Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening
Prospective function	Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening
Developability	Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening
Tested design count	Four therapeutic targets; ten Fc-fusion antibodies characterized in reported campaign
Success criterion	Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening
Independent validation	NR
Summary	Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening

Limitations and failure modes

Generation, ranking and display selection are intertwined.
reported affinities cannot be attributed to generator alone.
framework/design space constrained.

Deployment: Limited internal pilot after license review

Reproducibility, license, commercial use

Поле	Значение
Repo	https://github.com/TencentAI4S/tfold
License	Custom Tencent license; commercial partnership requested
Commercial-use status	Source available + weights/data; commercial review

Alternative tools: NR

Complementary: NR

Related: NR

Not directly comparable with

AlphaProteo, BindCraft, BoltzGen, Efficient evolution with ESM-1v, 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

Primary paper и полный список источников

tFold: epitope-specific de novo antibody 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: Antigen sequence, specified epitope and human framework

[MATRIX]Generated object: CDR sequences, candidate antibody structures and antibody–antigen complexes

[MATRIX]Фиксируется: NR. Изменяется: NR.

[MATRIX]De novo или parent-based: De novo CDRs on fixed human framework.

2. Архитектура в одном абзаце

[MATRIX]Sequence generator + tFold-Ab/tFold-Ag structure/complex models + structure-aware filters

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: CDR sequences, candidate antibody structures and antibody–antigen complexes. Structure generation: tFold-Ab and tFold-Ag integrated.

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): Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening

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): 3–5 prospective binding, epitope and developability evidence. Prospective binding: Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening. Function: Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening. Developability: Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening.

Generation, ranking and display selection are intertwined.
reported affinities cannot be attributed to generator alone.
framework/design space constrained.

15. Reproducibility

NRРепозиторий: https://github.com/TencentAI4S/tfold. Лицензия: Custom Tencent license; commercial partnership requested. Commit: NR, пока страница не имеет отдельного analysis YAML.

Модульная схема

Модуль	Класс модели	Вход	Выход	Обучается или frozen	Generator или evaluator
Sequence-based antibody generator	Sequence generator + tFold-Ab/tFold-Ag structure/complex models + structure-aware filters	NR	NR	NR	primary_generator
tFold-Ag binding/complex prediction + virtual-screen filters	NR	NR	NR	NR	integrated_scorer
tFold-Ab and tFold-Ag integrated	NR	NR	NR	NR	integrated_structure_predictor
Independent complex validation; phage display/SPR/competition assays	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
Sequence-based antibody generator	primary_generator	integrated	required	generation
tFold-Ag binding/complex prediction + virtual-screen filters	integrated_scorer	integrated	optional	post_hoc
tFold-Ab and tFold-Ag integrated	integrated_structure_predictor	integrated	optional	post_hoc
Independent complex validation; phage display/SPR/competition assays	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
Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening	Sequence generator + tFold-Ab/tFold-Ag structure/complex models + structure-aware filters	5. prospective functional validation	Four therapeutic targets; ten Fc-fusion antibodies characterized in reported campaign	NR	Best reported Kd: 3.22 nM Flu A, 81.6 nM PD-1, 0.045 nM PD-L1, 2.0 nM SC2RBD; after phage display screening	[MATRIX] landscape matrix 2026-08-13	Не читать как полный разбор primary paper.