Xikun Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xikun Huang has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Chaoqun Fei
Chinese Academy of Sciences
4 shared publications - Yangyang Li
Chinese Academy of Sciences
4 shared publications - Chuanqing Wang
Chinese Academy of Sciences
3 shared publications - Bingqing Liu
University of Chinese Academy of Science
2 shared publications - Baihui Tian
Changping Laboratory
1 shared publication - Di Wang
Changping Laboratory
1 shared publication
Latest Publications
- Open Access
High-Affinity Protein Binder Design via Flow Matching and In Silico Maturation
The de novo design of high-affinity protein binders remains a central challenge in protein engineering and therapeutic discovery. While deep generative models have advanced backbone generation and interface design, achieving picomolar or nanomolar binding affinities typically demands extensive experimental screening or iterative in vitro maturation. A general computational approach for the direct production of high-affinity binders has remained elusive. Here, we introduce an integrated framework that synergizes PPIFlow, a flow-matching-based generative model, with a novel in silico maturation strategy. PPIFlow employs a pairformer architecture to explicitly reason over pair-wise geometric and chemical interactions, modeling protein backbone rigid-body transformations as continuous flows. To bridge the affinity gap, we implement a dedicated in silico affinity maturation stage that combines interface rotamer enrichment with partial flow refinement to optimize energetic packing. This pipeline is further accelerated by AF3Score, a score-only adaptation of AlphaFold3 that enables high-fidelity and computationally efficient candidate prioritization. Across a diverse set of therapeutic targets, this synergistic approach consistently produces picomolar and nanomolar affinity binders without experimental affinity maturation. Notably, the framework proves highly effective for the de novo design of single-domain antibodies (VHHs), producing sub-nanomolar binders across multiple targets. These results establish that coupling robust backbone generation with focused in silico maturation renders the purely computational design of high-affinity binders feasible.
Open AccessHigh-Affinity Protein Binder Design via Flow Matching and In Silico Maturation
The de novo design of high-affinity protein binders remains a central challenge in protein engineering and therapeutic discovery. While deep generative models have advanced backbone generation and interface design, achieving picomolar or nanomolar binding affinities typically demands extensive experimental screening or iterative in vitro maturation. A general computational approach for the direct production of high-affinity binders has remained elusive. Here, we introduce an integrated framework that synergizes PPIFlow, a flow-matching-based generative model, with a novel in silico maturation strategy. PPIFlow employs a pairformer architecture to explicitly reason over pair-wise geometric and chemical interactions, modeling protein backbone rigid-body transformations as continuous flows. To bridge the affinity gap, we implement a dedicated in silico affinity maturation stage that combines interface rotamer enrichment with partial flow refinement to optimize energetic packing. This pipeline is further accelerated by AF3Score, a score-only adaptation of AlphaFold3 that enables high-fidelity and computationally efficient candidate prioritization. Across a diverse set of therapeutic targets, this synergistic approach consistently produces picomolar and nanomolar affinity binders without experimental affinity maturation. Notably, the framework proves highly effective for the de novo design of single-domain antibodies (VHHs), producing sub-nanomolar binders across multiple targets. These results establish that coupling robust backbone generation with focused in silico maturation renders the purely computational design of high-affinity binders feasible.
- 20 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/CAA Journal of Automatica Sinica
- Vol. 12 (9) , 1920-1937
Unsupervised Dynamic Discrete Structure Learning: A Geometric Evolution Method
Revealing the latent low-dimensional geometric structure of high-dimensional data is a crucial task in unsupervised representation learning. Traditional manifold learning, as a typical method for discovering latent geometric structures, has provided important nonlinear insight for the theoretical development of unsupervised representation learning. However, due to the shallow learning mechanism of the existing methods, they can only exploit the simple geometric structure embedded in the initial data, such as the local linear structure. Traditional manifold learning methods are fairly limited in mining higher-order non-linear geometric information, which is also crucial for the development of unsupervised representation learning. To address the abovementioned limitations, this paper proposes a novel dynamic geometric structure learning model (DGSL) to explore the true latent nonlinear geometric structure. Specifically, by mathematically analysing the reconstruction loss function of manifold learning, we first provide universal geometric relational function between the curvature and the non-Euclidean metric of the initial data. Then, we leverage geometric flow to design a deeply iterative learning model to optimize this relational function. Our method can be viewed as a general-purpose algorithm for mining latent geometric structures, which can enhance the performance of geometric representation methods. Experimentally, we perform a set of representation learning tasks on several datasets. The experimental results show that our proposed method is superior to traditional methods.
- 20 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/CAA Journal of Automatica Sinica
- Vol. 12 (9) , 1920-1937
Unsupervised Dynamic Discrete Structure Learning: A Geometric Evolution Method
Revealing the latent low-dimensional geometric structure of high-dimensional data is a crucial task in unsupervised representation learning. Traditional manifold learning, as a typical method for discovering latent geometric structures, has provided important nonlinear insight for the theoretical development of unsupervised representation learning. However, due to the shallow learning mechanism of the existing methods, they can only exploit the simple geometric structure embedded in the initial data, such as the local linear structure. Traditional manifold learning methods are fairly limited in mining higher-order non-linear geometric information, which is also crucial for the development of unsupervised representation learning. To address the abovementioned limitations, this paper proposes a novel dynamic geometric structure learning model (DGSL) to explore the true latent nonlinear geometric structure. Specifically, by mathematically analysing the reconstruction loss function of manifold learning, we first provide universal geometric relational function between the curvature and the non-Euclidean metric of the initial data. Then, we leverage geometric flow to design a deeply iterative learning model to optimize this relational function. Our method can be viewed as a general-purpose algorithm for mining latent geometric structures, which can enhance the performance of geometric representation methods. Experimentally, we perform a set of representation learning tasks on several datasets. The experimental results show that our proposed method is superior to traditional methods.
- Open Access
Link-aware link prediction over temporal graph by pattern recognition
A temporal graph can be considered as a stream of links, each of which represents an interaction between two nodes at a certain time. On temporal graphs, link prediction is a common task, which aims to answer whether the query link is true or not. To do this task, previous methods usually focus on the learning of representations of the two nodes in the query link. We point out that the learned representation by their models may encode too much information with side effects for link prediction because they have not utilized the information of the query link, i.e., they are link-unaware. Based on this observation, we propose a link-aware model: historical links and the query link are input together into the following model layers to distinguish whether this input implies a reasonable pattern that ends with the query link. During this process, we focus on the modeling of link evolution patterns rather than node representations. Experiments on six datasets show that our model achieves strong performances compared with state-of-the-art baselines, and the results of link prediction are interpretable. The code and datasets are available on the project website: https://github.com/lbq8942/TGACN.
Open AccessLink-aware link prediction over temporal graph by pattern recognition
A temporal graph can be considered as a stream of links, each of which represents an interaction between two nodes at a certain time. On temporal graphs, link prediction is a common task, which aims to answer whether the query link is true or not. To do this task, previous methods usually focus on the learning of representations of the two nodes in the query link. We point out that the learned representation by their models may encode too much information with side effects for link prediction because they have not utilized the information of the query link, i.e., they are link-unaware. Based on this observation, we propose a link-aware model: historical links and the query link are input together into the following model layers to distinguish whether this input implies a reasonable pattern that ends with the query link. During this process, we focus on the modeling of link evolution patterns rather than node representations. Experiments on six datasets show that our model achieves strong performances compared with state-of-the-art baselines, and the results of link prediction are interpretable. The code and datasets are available on the project website: https://github.com/lbq8942/TGACN.
- 15 November 2023
- Published by Springer Nature
- p. 395-402
Link Prediction Based on Pairwise Proximity Preserving Graph Neural Networks
- 15 November 2023
- Published by Springer Nature
- p. 395-402
Link Prediction Based on Pairwise Proximity Preserving Graph Neural Networks
- 15 November 2023
- Published by Springer Nature
- p. 283-297
Nonlinear Manifold Learning via Graph Curvature
- 15 November 2023
- Published by Springer Nature
- p. 283-297
Nonlinear Manifold Learning via Graph Curvature
- 19 July 2023
- Published by Springer Nature
- p. 325-337
Link-Aware Link Prediction over Temporal Graph by Pattern Recognition
- 19 July 2023
- Published by Springer Nature
- p. 325-337
Link-Aware Link Prediction over Temporal Graph by Pattern Recognition
- 19 July 2022
- Published by Springer Nature
- p. 450-465
Labeled Knowledge-Based Decision Making with Assumption-Based Argumentation
- 19 July 2022
- Published by Springer Nature
- p. 450-465
Labeled Knowledge-Based Decision Making with Assumption-Based Argumentation