IntFold: A Controllable Foundation Model for General and Specialized Biomolecular Structure Prediction

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: The IntFold Team, Qiao, Leon, Bai, Wayne, Yan, He, Liu, Gary, Xi, Nova, Zhang, Xiang, Sun, Siqi
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912464609738752
author The IntFold Team
Qiao, Leon
Bai, Wayne
Yan, He
Liu, Gary
Xi, Nova
Zhang, Xiang
Sun, Siqi
author_facet The IntFold Team
Qiao, Leon
Bai, Wayne
Yan, He
Liu, Gary
Xi, Nova
Zhang, Xiang
Sun, Siqi
contents We introduce IntFold, a controllable foundation model for general and specialized biomolecular structure prediction. Utilizing a high-performance custom attention kernel, IntFold achieves accuracy comparable to the state-of-the-art AlphaFold 3 on a comprehensive benchmark of diverse biomolecular structures, while also significantly outperforming other leading all-atom prediction approaches. The model's key innovation is its controllability, enabling downstream applications critical for drug screening and design. Through specialized adapters, it can be precisely guided to predict complex allosteric states, apply user-defined structural constraints, and estimate binding affinity. Furthermore, we present a training-free, similarity-based method for ranking predictions that improves success rates in a model-agnostic manner. This report details these advancements and shares insights from the training and development of this large-scale model.
format Preprint
id arxiv_https___arxiv_org_abs_2507_02025
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle IntFold: A Controllable Foundation Model for General and Specialized Biomolecular Structure Prediction
The IntFold Team
Qiao, Leon
Bai, Wayne
Yan, He
Liu, Gary
Xi, Nova
Zhang, Xiang
Sun, Siqi
Biomolecules
We introduce IntFold, a controllable foundation model for general and specialized biomolecular structure prediction. Utilizing a high-performance custom attention kernel, IntFold achieves accuracy comparable to the state-of-the-art AlphaFold 3 on a comprehensive benchmark of diverse biomolecular structures, while also significantly outperforming other leading all-atom prediction approaches. The model's key innovation is its controllability, enabling downstream applications critical for drug screening and design. Through specialized adapters, it can be precisely guided to predict complex allosteric states, apply user-defined structural constraints, and estimate binding affinity. Furthermore, we present a training-free, similarity-based method for ranking predictions that improves success rates in a model-agnostic manner. This report details these advancements and shares insights from the training and development of this large-scale model.
title IntFold: A Controllable Foundation Model for General and Specialized Biomolecular Structure Prediction
topic Biomolecules
url https://arxiv.org/abs/2507.02025