SCSen Cao

Sen Cao

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Sen Cao has a total of 33 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    HighFold-Lasso: Topology-Guided Prediction of Lasso Peptide Structures Using AlphaFold3

    by Sen Cao, Chengyun Zhang, Hanjie Yang, Yutong Ge, Hongli Hou, Ning Zhu, Yaling Wu, Jisong Mo, Juan Guo, Chongyang Li, et al.

    Lasso peptides are a class of ribosomally synthesized and post-translationally modified peptides (RiPPs) characterized by a mechanically interlocked topology, in which the C-terminal tail is threaded through an N-terminal macrolactam ring formed by an isopeptide bond. This unique topology often confers high thermal stability and resistance to proteolytic degradation. However, the isopeptide linkage, threaded topology, and class-specific disulfide connectivity make lasso peptide structures difficult to reconstruct reliably using general-purpose structure prediction models. Here, we developed HighFold-Lasso, which combines custom covalent connectivity with lasso scaffold-guided adaptive low-noise initialization to enable conditional structure generation guided by lasso conformational priors. On a test set of 52 experimentally determined structures, HighFold-Lasso achieved a 100% lasso-topology success rate, with mean RMSD Cα of 1.942 Å for the ring–loop region and 2.375 Å for the full peptide. The framework supports the modeling of the four classical lasso peptide classes (I–IV) and can represent noncanonical amino acids, providing computational support for structure–function studies and molecular engineering of lasso peptides.

  • Open Access

    HighFold4: extending AlphaFold3 to accurate cyclic peptide conformation prediction via custom chemical connectivity

    by Chengyun Zhang, Wentong Wang, Renjie Zhu, Sen Cao, Yiqi Xu, Ning Zhu, Yaling Wu, Jingjing Guo, Hongliang Duan

    While AlphaFold3 has revolutionized protein structure prediction and supports noncanonical amino acids, its architecture always fails to reliably generate the closed-ring topologies characteristic of cyclic peptides. Existing adaptations, such as imposing distance constraints via an offset matrix, enforce ring geometry but cannot specify the chemical identity of the cyclization bond, leading to a restrictive bias toward amide- or disulfide-linked macrocycles. Here, we present HighFold4, a framework that adapts AlphaFold3 to explicitly incorporate user-defined chemical connectivity between residues, thereby enabling both topological closure and bond-specific cyclization. Without retraining the base model, HighFold4 achieves accurate, chemistry-aware prediction of diverse cyclic peptide conformations, as validated on 179 structures. This work established a new paradigm for the conformation construction of macrocyclic peptides with tailored ring geometry and linkage chemistry, significantly expanding the utility of deep learning in peptide-based drug discovery.


  • HighMorph: De Novo Cyclic Peptide Sequence Design via Protein–Protein Interaction Recapitulation

    by Minhui Lan, Chengyun Zhang, Wentong Wang, Haomeng Hu, Huitian Lin, Sen Cao, Jingjing Guo, Hongliang Duan

    Cyclic peptides have emerged as a compelling class of bioactive scaffolds, but de novo design of target-binding cyclic peptides from protein structures remains challenging. Here, we present HighMorph, an interaction-guided framework that combines protein–protein interaction information with artificial intelligence for rational cyclic peptide design. HighMorph integrates Monte Carlo tree search with a Transformer-based policy-value network to efficiently explore cyclic peptide sequence space, while incorporating explicit atomic-level hydrogen bond constraints extracted from reference protein–protein complexes to guide sequence optimization. The framework is systematically validated on two clinically relevant targets, programmed death-ligand 1 (PD-L1) and kallikrein-related peptidase 4 (KLK4). Notably, 33.3% and 40% of the generated candidates are active against PD-L1 and KLK4, respectively, with active cyclic peptides exhibiting micromolar binding affinities (approximately 10–6 M). These results validate our approach for cyclic peptide design. Additionally, interaction analysis provides insights for developing therapeutics targeting challenging protein interfaces.

  • HighFold-MeD2: An Enhanced Boltz-2 Model for Accurate Structure Prediction of N-Methylated and d -Amino Acid Cyclic Peptides

    by Zhigang Cao, Wentong Wang, Sen Cao, Chengyun Zhang, Jingjing Guo, Hongliang Duan

    Cyclic peptides containing backbone N-methylated amino acids (BNMeAAs) and d-amino acids (d-AAs) exhibit exceptional metabolic stability, membrane permeability, and oral bioavailability, representing a pivotal middle space scaffold between biologics and s small-molecule drugs. To optimize pharmacokinetic and physicochemical properties, various non-canonical amino acids(NCAAs) are frequently introduced; however, these modifications pose significant challenges for three-dimensional structure prediction. While Rosetta simple_cycpep_predict (SCP) can generate high-accuracy structures across a broad chemical space based on energy functions, it requires extensive conformational sampling, resulting in prohibitive computational costs. Recently, deep learning approaches such as AlphaFold3, Boltz-2, HighFold2, and NCPepFold have advanced cyclic peptide prediction but offer limited support for BNMeAAs and d-AAs. In previous work, HighFold-MeD was proposed, which distilled massive conformational data generated by Rosetta SCP into the AlphaFold2 framework, significantly accelerating prediction for modified cyclic peptides. Building upon this, the present study introduces HighFold-MeD2, an enhanced model based on Boltz-2. By leveraging the sequence, modification information, and cyclic constraints as inputs, HighFold-MeD2 utilizes the Pairformer-Diffusion architecture of Boltz-2 to learn the generation of all-atom coordinates from random initialization to Rosetta target structures. The predicted structures are subsequently refined via Amber force field energy minimization to resolve local steric clashes. Unlike its AlphaFold2-based predecessor, which relied on labor-intensive hard-coding for non-natural amino acids, HighFold-MeD2 achieves seamless integration through universal Chemical Component Dictionary (CCD) representations. It not only demonstrates superior prediction precision over HighFold-MeD and state-of-the-art baselines but also validates the feasibility of utilizing CADD-generated conformations to fine-tune all-atom diffusion models, establishing a highly accessible and extensible paradigm for cyclic peptide drug design.

  • High-PepBinder: A pLM-Guided Latent Diffusion Framework for Affinity-Aware Target-Specific Peptide Design

    by Qingyi Mao, Silong Zhai, Sen Cao, Renjie Zhu, Wen Xu, Chengyun Zhang, Ning Zhu, Jingjing Guo, Hongliang Duan

    Peptides, as therapeutic molecules, offer unique advantages in targeting complex protein surfaces, yet their rational design remains limited by the vastness of the sequence space and the constraints of traditional approaches. Here, we propose High-PepBinder, a sequence-only conditional diffusion framework for target-specific peptide generation. Guided by the target protein sequence, High-PepBinder adopts a dual encoder architecture that integrates protein language models (pLMs) with the diffusion model. This approach cascades the peptide generation model with an affinity classifier and enables the generation process to capture affinity-related features of the peptides through lightweight joint optimization. Due to the scarcity of protein-peptide affinity data, we constructed PepPBA, to our knowledge, the most comprehensive dataset to date, and established a structure- and physics-based screening pipeline to prioritize top candidates. Results show that High-PepBinder demonstrates competitive performance across multiple peptide generation and binding-related evaluation tasks. For representative targets, including KEAP1, XIAP and EGFR, the generated peptides preserve key binding geometries and interface patterns of reference peptides in predicted complexes, while maintaining sequence diversity and favorable predicted properties. Overall, High-PepBinder contributes toward a general and sequence-only strategy for peptide design, offering a computational framework for expanding peptide discovery against challenging targets.


  • HighRelax: Physics-Based Refinement of Deep Learning Protein Predictions with Noncanonical Amino Acids

    by Sen Cao, Chengyun Zhang, Ning Zhu, Chongyang Li, Qingyi Mao, Zhigang Cao, Yutong Ge, Yaling Wu, Juan Guo, Qiang Cao, et al.

    Noncanonical amino acids (NCAAs) have emerged as essential building blocks in protein engineering and peptide drug development owing to their advantages in enhancing metabolic stability, membrane permeability, and resistance to proteolytic degradation. Accurate construction of 3D protein structures containing NCAAs is crucial for elucidating their functions, understanding molecular interactions, and enabling rational design. However, integrating NCAAs into state-of-the-art protein structure prediction frameworks─such as AlphaFold3─often results in chirality violations, steric clashes, and local geometric distortions. These issues likely reflect limited parametrization of nonstandard residues within current models. To address these challenges, we expanded the AMBER force field covering 139 NCAAs, and we developed an enhanced Amber-relax protocol named HighRelax. Unlike conventional workflows that are restricted to linear peptides composed of canonical amino acids, HighRelax is compatible with complex systems containing NCAAs and cyclic peptides and can be seamlessly integrated with structures generated by state-of-the-art models such as AlphaFold3. Our results demonstrate that HighRelax effectively reduces steric clashes, restores residue chirality, and improves overall structural quality. This method provides a general postprocessing strategy for refining NCAA-containing structures, facilitating their applications in molecular simulation, peptide drug design, and protein engineering.


  • Open Access

    High-PepBinder: A pLM-Guided Latent Diffusion Framework for Affinity-Aware Target-Specific Peptide Design

    by Qingyi Mao, Silong Zhai, Sen Cao, Renjie Zhu, Wen Xu, Chengyun Zhang, Ning Zhu, Jingjing Guo, Hongliang Duan

    Peptides, as therapeutic molecules, offer unique advantages in targeting complex protein surfaces, yet their rational design remains limited by the vastness of the sequence space and the constraints of traditional approaches. Here, we propose High-PepBinder, a sequence-only conditional diffusion framework for target-specific peptide generation. Guided by the target protein sequence, High-PepBinder adopts a dual encoder architecture that integrates protein language models (pLMs) with the diffusion model. This approach cascades the peptide generation model with an affinity classifier and enables the generation process to capture affinity-related features of the peptides through lightweight joint optimization. Due to the scarcity of protein-peptide affinity data, we constructed PepPBA, to our knowledge the most comprehensive dataset to date, and established a structure- and physics-based screening pipeline to prioritize top candidates. Results show that High-PepBinder demonstrates competitive performance across multiple peptide generation and affinity-related tasks. For representative targets, including KEAP1, XIAP, and EGFR, the generated peptides preserve key binding geometries and interface patterns of reference peptides in predicted complexes, while maintaining sequence diversity and favorable predicted properties. Overall, High-PepBinder contributes toward a general and sequence-only strategy for peptide design, offering a computational framework for expanding peptide discovery against challenging targets.


  • Open Access

    Accurate structure prediction of cyclic peptides containing unnatural amino acids using HighFold3

    by Sen Cao, Cheng Zhu, Qingyi Mao, Jingjing Guo, Ning Zhu, Hongliang Duan

    Cyclic peptides have emerged as a research hotspot in drug development in recent years due to their excellent stability, specificity, and cell penetration. However, existing computational models face challenges in accurately predicting the three-dimensional structures of cyclic peptides containing unnatural amino acids (unAAs), thereby limiting their drug design. The release of AlphaFold 3 has significantly enhanced the modeling capability of biomolecular complexes and enabled the inclusion of unAAs through definitions provided by the Chemical Component Dictionary (CCD). Nevertheless, its training data reliance limits its ability to accurately predict cyclic peptide structures, failing to meet the demand for precise cyclic peptide structure prediction. Based on the AlphaFold 3 framework, we developed HighFold3 by introducing the Cyclic Position Offset Encoding Matrix (CycPOEM). HighFold3 comprises two submodels: HighFold3-Linear and HighFold3–Cyclic, designed for predicting the structures of linear and cyclic peptides, respectively. Our results demonstrate that HighFold3 outperforms existing models (HighFold, HighFold2, CyclicBoltz1, NCPepFold, CABS-flex, ESMFold, and HelixFold) in cyclic peptide structure prediction. It achieves atomic-level precision in predicting cyclic peptide monomers while demonstrating enhanced accuracy and generalization capability for cyclic peptide complexes containing unAAs. This offers unprecedented technical support for the structural design and optimization of cyclic peptide–based therapeutics.


  • Open Access

    Predicting the structures of cyclic peptides containing unnatural amino acids by HighFold2

    by Cheng Zhu, Sen Cao, Tianfeng Shang, Jingjing Guo, An Su, Chengxi Li, Hongliang Duan

    Cyclic peptides containing unnatural amino acids possess many excellent properties and have become promising candidates in drug discovery. Therefore, accurately predicting the 3D structures of cyclic peptides containing unnatural residues will significantly advance the development of cyclic peptide-based therapeutics. Although deep learning-based structural prediction models have made tremendous progress, these models still cannot predict the structures of cyclic peptides containing unnatural amino acids. To address this gap, we introduce a novel model, HighFold2, built upon the AlphaFold-Multimer framework. HighFold2 first extends the pre-defined rigid groups and their initial atomic coordinates from natural amino acids to unnatural amino acids, thus enabling structural prediction for these residues. Then, it incorporates an additional neural network to characterize the atom-level features of peptides, allowing for multi-scale modeling of peptide molecules while enabling the distinction between various unnatural amino acids. Besides, HighFold2 constructs a relative position encoding matrix for cyclic peptides based on different cyclization constraints. Except for training using spatial structures with unnatural amino acids, HighFold2 also parameterizes the unnatural amino acids to relax the predicted structure by energy minimization for clash elimination. Extensive empirical experiments demonstrate that HighFold2 can accurately predict the 3D structures of cyclic peptide monomers containing unnatural amino acids and their complexes with proteins, with the median RMSD for Cα reaching 1.891 Å. All these results indicate the effectiveness of HighFold2, representing a significant advancement in cyclic peptide-based drug discovery.