FABind

Abstract

Modeling the interaction between proteins and ligands and accurately predicting their binding structures is a critical yet challenging task in drug discovery. Recent advancements in deep learning have shown promise in addressing this challenge, with sampling-based and regression-based methods emerging as two prominent approaches. However, these methods have notable limitations. Sampling-based methods often suffer from low efficiency due to the need for generating multiple candidate structures for selection. On the other hand, regression-based methods offer fast predictions but may experience decreased accuracy. Additionally, the variation in protein sizes often requires external modules for selecting suitable binding pockets, further impacting efficiency. In this work, we propose FABind, an end-to-end model that combines pocket prediction and docking to achieve accurate and fast protein-ligand binding. FABind incorporates a unique ligand-informed pocket prediction module, which is also leveraged for docking pose estimation. The model further enhances the docking process by incrementally integrating the predicted pocket to optimize protein-ligand binding, reducing discrepancies between training and inference. Through extensive experiments on benchmark datasets, our proposed FABind demonstrates strong advantages in terms of effectiveness and efficiency compared to existing methods.

Demo (PDBBind v2020)

PDB ID	RMSD

Transparency:

Red: ground truth ligand pose. Green: ligand pose predicted by FABind.

Methods

Framework

We divide the task into pocket prediction and docking pose prediction:
- Pocket prediction: take the whole protein and the ligand as input and predicts the coordinates of the pocket center, where the ligand is randomly placed at the center of the protein. After determining the pocket center, a pocket is defined as a set of amino acids within a fixed radius around the center.
- Docking pose prediction: the ligand is moved to the pocket center and the ligand-pocket pair iteratively goes through the FABind layers to obtain the final pose prediction.
Advantages:
- Unified network for pocket prediction and docking pose prediction.
- End-to-end framework. Input rigid protein and random molecule conformation, then output the predicted pose.
- Scheduled sampling. The predicted pockets are gradually used in docking pose prediction.

FABind Layer

FABind layer consists of the following three parts:
- Independent message passing. The independent encoder first passes messages inside the protein and ligand to update node embeddings and coordinates.
- Cross-attention update. This block operates to exchange information across each node between protein and ligand and updates pair embeddings accordingly.
- Interfacial message passing. This layer focuses on the contact surface and attentively updates coordinates and representations for nodes on the contact surface.

Results

Flexible Blind Self-docking Performance on All Receptors

Across most metrics, FABind ranking either as the best or second best. FABind exceeds the DiffDock with 10 sampled poses. Although FABind may not achieve the best performance in terms of the <2Å metric, it demonstrates exceptional proficiency with a mean RMSD score of 6.4Å.
For efficiency, the inference time cost of FABind ranks second lowest. Notably, when compared to DiffDock with 10 sampled poses, our FABind achieves better docking performance while being over 170 times faster.

Flexible Blind Self-docking Performance on Unseen Receptors

To perform a more rigorous evaluation, we apply an additional filtering step to exclude samples whose UniProt IDs of the proteins are not contained in the data that is seen during training and validation. The resulting subset, consisting of 144 complexes, was then used to evaluate the performance of FABind. The results demonstrate that FABind surpasses other deep learning and traditional methods by a significant margin across all evaluation metrics. These findings strongly indicate the robust generalization capability of FABind.

Flexible Blind Self-docking Performance on Apo-proteins

In real-world applications, docking is often performed on apo or holo-structures that are bound to different ligands. A new benchmark is to combine the crystal complex structures of PDBBind with protein structures predicted by ESMFold. In order to validate the efficacy of our FABind in the apo-structure docking scenario, we also evaluated its performance under the same settings with DiffDock. FABind outperforms DiffDock, achieving an RMSD of less than 2Å on 24.9% of the complexes generated by ESMFold.

Pocket Prediction Analysis

Removing the center constraint (classification only) significantly hurts the prediction performance.
The integration of ligand information into the pocket prediction task enhances performance and efficiency.
Combined with center constraint, even without incorporating ligand information, our pocket prediction performance outperforms TankBind and E3Bind.

Citation


@inproceedings{
  pei2023fabind,
  title={{FAB}ind: Fast and Accurate Protein-Ligand Binding},
  author={Qizhi Pei and Kaiyuan Gao and Lijun Wu and Jinhua Zhu and Yingce Xia and Shufang Xie and Tao Qin and Kun He and Tie-Yan Liu and Rui Yan},
  booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
  year={2023},
  url={https://openreview.net/forum?id=PnWakgg1RL}
}

FABind: Fast and Accurate Protein-Ligand Binding