Haiyan Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Haiyan Liu has a total of 213 co-authors. Here are the top co-authors by number of shared publications:
- Quan Chen
Department of Rheumatology and Immunology, The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, Hefei National Research Center for Physical Sciences at the Microscale, Center for Advanced Interdisciplinary Science and Biomedicine of IHM, University of Science and Technology of China, Hefei, Anhui, China.
20 shared publications Jiahai Zhang
MOE Key Laboratory for Membraneless Organelles and Cellular Dynamics, School of Life Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, China
8 shared publications- Yufeng Liu
Department of Rheumatology and Immunology, The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, Hefei National Research Center for Physical Sciences at the Microscale, Center for Advanced Interdisciplinary Science and Biomedicine of IHM, University of Science and Technology of China, Hefei, Anhui, China
6 shared publications Chenchen Wang
MOE Key Laboratory for Membraneless Organelles and Cellular Dynamics, School of Life Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, Anhui 230027, China
5 shared publicationsFudong Li
MOE Key Laboratory for Membraneless Organelles and Cellular Dynamics, School of Life Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, China
4 shared publicationsLinghui Chen
Oristruct Biotech Co., Ltd, Hefei, Anhui 230026, China
4 shared publications
Latest Publications
- 23 April 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (9) , 4199-4200
Advancing Reproducibility and Open Data in Theoretical and Computational Chemistry
- 23 April 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (9) , 4199-4200
Advancing Reproducibility and Open Data in Theoretical and Computational Chemistry
- Open Access
- 19 January 2026
- Published by Elsevier in New Biotechnology
- Vol. 92, 51-61
Zero-shot deep learning with multi-objective optimization improves thermostability of zearalenone hydrolase and xylanase
- 19 January 2026
- Published by Elsevier in New Biotechnology
- Vol. 92, 51-61
Open AccessZero-shot deep learning with multi-objective optimization improves thermostability of zearalenone hydrolase and xylanase
- Open Access
- 1 January 2026
- Published by China Engineering Science Press Co. Ltd. in Vita
Designing substrate-specific metalloproteases
In a recent study published in Vita, Qu et al. leverage ProteusFlow, a flow-based backbone generation method, to design substrate-encapsulating metalloproteases targeting three distinct cleavage sites in the amyloid-β peptide, achieving programmable substrate recognition and high catalytic efficiency in a subset of designs.
- 1 January 2026
- Published by China Engineering Science Press Co. Ltd. in Vita
Open AccessDesigning substrate-specific metalloproteases
In a recent study published in Vita, Qu et al. leverage ProteusFlow, a flow-based backbone generation method, to design substrate-encapsulating metalloproteases targeting three distinct cleavage sites in the amyloid-β peptide, achieving programmable substrate recognition and high catalytic efficiency in a subset of designs.
- Open Access
- 19 November 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 10177
Enhancing functional proteins through multimodal inverse folding with ABACUS-T
Structure-based sequence redesign or inverse folding can significantly enhance structural stability but often compromises functional activity when performed using existing models. Here, we introduce ABACUS-T, a multimodal inverse folding model that improves precision and minimizes functional loss. ABACUS-T unifies several important features into one framework: detailed atomic sidechains and ligand interactions, a pre-trained protein language model, multiple backbone conformational states, and evolutionary information from multiple sequence alignment (MSA). Redesigned proteins show notable improvements: an allose binding protein achieves 17-fold higher affinity while retaining conformational change; redesigned endo-1,4-β-xylanase and TEM β-lactamase maintain or surpass wild-type activity; and OXA β-lactamase gains altered substrate selectivity. All achieve substantially increase thermostability (∆Tm ≥ 10 °C). In each test case, these enhancements are achieved by testing only a few sequences, each containing dozens of simultaneously mutated residues. ABACUS-T thus offers a promising tool for reengineering functional proteins in biotechnological applications.
- 19 November 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 10177
Open AccessEnhancing functional proteins through multimodal inverse folding with ABACUS-T
Structure-based sequence redesign or inverse folding can significantly enhance structural stability but often compromises functional activity when performed using existing models. Here, we introduce ABACUS-T, a multimodal inverse folding model that improves precision and minimizes functional loss. ABACUS-T unifies several important features into one framework: detailed atomic sidechains and ligand interactions, a pre-trained protein language model, multiple backbone conformational states, and evolutionary information from multiple sequence alignment (MSA). Redesigned proteins show notable improvements: an allose binding protein achieves 17-fold higher affinity while retaining conformational change; redesigned endo-1,4-β-xylanase and TEM β-lactamase maintain or surpass wild-type activity; and OXA β-lactamase gains altered substrate selectivity. All achieve substantially increase thermostability (∆Tm ≥ 10 °C). In each test case, these enhancements are achieved by testing only a few sequences, each containing dozens of simultaneously mutated residues. ABACUS-T thus offers a promising tool for reengineering functional proteins in biotechnological applications.
- Open Access
- 16 July 2025
- Published by Oxford University Press (OUP) in National Science Review
- Vol. 12 (11) , nwaf290
Designing flexible protein structures and sampling protein conformations with a unified model using vector quantization and diffusion
Conformational dynamics are often critical for protein functions. There is strong interest in deep learning models to predict the conformational distributions of proteins or design protein structures that can host rich conformational dynamics. Here we report PVQD (Protein Vector Quantization and Diffusion), a method using a vector-quantized auto-encoder to learn protein backbone latent representations and latent-space diffusion for backbone generation and for conformation sampling conditioned on native sequences. Comparisons show PVQD generates backbones with natural-like compositions of secondary structures, loop lengths, and domain sizes. In sampling conformations of natural proteins, PVQD better reproduces experimental structural variations in benchmark proteins than existing approaches. For K-Ras, KaiB, 4.1G CTD, and D-allose binding proteins, PVQD captures sequence-dependent effects on functional conformational dynamics. Thus, the latent space diffusion approach forms a valuable framework that can unify the prediction and design of protein structures with the capability of modeling conformational dynamics.
- 16 July 2025
- Published by Oxford University Press (OUP) in National Science Review
- Vol. 12 (11) , nwaf290
Open AccessDesigning flexible protein structures and sampling protein conformations with a unified model using vector quantization and diffusion
Conformational dynamics are often critical for protein functions. There is strong interest in deep learning models to predict the conformational distributions of proteins or design protein structures that can host rich conformational dynamics. Here we report PVQD (Protein Vector Quantization and Diffusion), a method using a vector-quantized auto-encoder to learn protein backbone latent representations and latent-space diffusion for backbone generation and for conformation sampling conditioned on native sequences. Comparisons show PVQD generates backbones with natural-like compositions of secondary structures, loop lengths, and domain sizes. In sampling conformations of natural proteins, PVQD better reproduces experimental structural variations in benchmark proteins than existing approaches. For K-Ras, KaiB, 4.1G CTD, and D-allose binding proteins, PVQD captures sequence-dependent effects on functional conformational dynamics. Thus, the latent space diffusion approach forms a valuable framework that can unify the prediction and design of protein structures with the capability of modeling conformational dynamics.
- Open Access
- 27 February 2025
- Published by Oxford University Press (OUP) in Nucleic Acids Research
- Vol. 53 (5)
Decoding and reengineering the promoter specificity of T7-like RNA polymerases based on phage genome sequences
The single subunit RNA polymerases (ssRNAPs) of bacteriophages are highly interesting targets for the prediction and engineering of specific protein–DNA interactions. Despite extensive existing studies focusing on particular ssRNAPs such as the T7 RNAP, few rules governing the protein–DNA sequence covariations across diverse ssRNAPs and their cognate promoters are clearly known. Here, aiming to reveal such rules, we comprehensively mined promoters of various categories of ssRNAPs from phage genomes. For T7-like RNAPs, direct coupling analyses of the predicted set of RNAP–promoter pairs revealed that the interaction specificity was dominantly encoded by the amino acid and nucleotide residues at only a few key positions. The covariations between the amino acid and the nucleotide residues at these positions were summarized into a sparsely connected network. Using experimentally verified connections in this network, we designed a set of orthogonal T7 RNAP–promoter variants that showed more stringent orthogonality than previously reported sets. We further designed and experimentally verified variants with novel interactions. These results provided guidance for engineering novel RNAP–promoter pairs for synthetic biology or other applications. Our study also demonstrated the use of comprehensive genome mining in combination with sequence covariation analysis in the prediction and engineering of specific protein–DNA interactions.
- 27 February 2025
- Published by Oxford University Press (OUP) in Nucleic Acids Research
- Vol. 53 (5)
Open AccessDecoding and reengineering the promoter specificity of T7-like RNA polymerases based on phage genome sequences
The single subunit RNA polymerases (ssRNAPs) of bacteriophages are highly interesting targets for the prediction and engineering of specific protein–DNA interactions. Despite extensive existing studies focusing on particular ssRNAPs such as the T7 RNAP, few rules governing the protein–DNA sequence covariations across diverse ssRNAPs and their cognate promoters are clearly known. Here, aiming to reveal such rules, we comprehensively mined promoters of various categories of ssRNAPs from phage genomes. For T7-like RNAPs, direct coupling analyses of the predicted set of RNAP–promoter pairs revealed that the interaction specificity was dominantly encoded by the amino acid and nucleotide residues at only a few key positions. The covariations between the amino acid and the nucleotide residues at these positions were summarized into a sparsely connected network. Using experimentally verified connections in this network, we designed a set of orthogonal T7 RNAP–promoter variants that showed more stringent orthogonality than previously reported sets. We further designed and experimentally verified variants with novel interactions. These results provided guidance for engineering novel RNAP–promoter pairs for synthetic biology or other applications. Our study also demonstrated the use of comprehensive genome mining in combination with sequence covariation analysis in the prediction and engineering of specific protein–DNA interactions.
- 9 October 2024
- Published by Springer Nature in Nature Methods
- Vol. 21 (11) , 2107-2116
De novo protein design with a denoising diffusion network independent of pretrained structure prediction models
The recent success of RFdiffusion, a method for protein structure design with a denoising diffusion probabilistic model, has relied on fine-tuning the RoseTTAFold structure prediction network for protein backbone denoising. Here, we introduce SCUBA-diffusion (SCUBA-D), a protein backbone denoising diffusion probabilistic model freshly trained by considering co-diffusion of sequence representation to enhance model regularization and adversarial losses to minimize data-out-of-distribution errors. While matching the performance of the pretrained RoseTTAFold-based RFdiffusion in generating experimentally realizable protein structures, SCUBA-D readily generates protein structures with not-yet-observed overall folds that are different from those predictable with RoseTTAFold. The accuracy of SCUBA-D was confirmed by the X-ray structures of 16 designed proteins and a protein complex, and by experiments validating designed heme-binding proteins and Ras-binding proteins. Our work shows that deep generative models of images or texts can be fruitfully extended to complex physical objects like protein structures by addressing outstanding issues such as the data-out-of-distribution errors.
- 9 October 2024
- Published by Springer Nature in Nature Methods
- Vol. 21 (11) , 2107-2116
De novo protein design with a denoising diffusion network independent of pretrained structure prediction models
The recent success of RFdiffusion, a method for protein structure design with a denoising diffusion probabilistic model, has relied on fine-tuning the RoseTTAFold structure prediction network for protein backbone denoising. Here, we introduce SCUBA-diffusion (SCUBA-D), a protein backbone denoising diffusion probabilistic model freshly trained by considering co-diffusion of sequence representation to enhance model regularization and adversarial losses to minimize data-out-of-distribution errors. While matching the performance of the pretrained RoseTTAFold-based RFdiffusion in generating experimentally realizable protein structures, SCUBA-D readily generates protein structures with not-yet-observed overall folds that are different from those predictable with RoseTTAFold. The accuracy of SCUBA-D was confirmed by the X-ray structures of 16 designed proteins and a protein complex, and by experiments validating designed heme-binding proteins and Ras-binding proteins. Our work shows that deep generative models of images or texts can be fruitfully extended to complex physical objects like protein structures by addressing outstanding issues such as the data-out-of-distribution errors.
- Open Access
- 11 September 2024
- Published by Springer Nature in Cell Research
- Vol. 34 (12) , 820-821
Designing de novo D-protein binders
- 11 September 2024
- Published by Springer Nature in Cell Research
- Vol. 34 (12) , 820-821
Open AccessDesigning de novo D-protein binders
- 29 March 2024
- Published by American Chemical Society (ACS) in ACS Synthetic Biology
- Vol. 13 (4) , 1177-1190
Using Protein Design and Directed Evolution to Monomerize a Bright Near-Infrared Fluorescent Protein
The small ultrared fluorescent protein (smURFP) is a bright near-infrared (NIR) fluorescent protein (FP) that forms a dimer and binds its fluorescence chromophore, biliverdin, at its dimer interface. To engineer a monomeric NIR FP based on smURFP potentially more suitable for bioimaging, we employed protein design to extend the protein backbone with a new segment of two helices that shield the original dimer interface while covering the biliverdin binding pocket in place of the second chain in the original dimer. We experimentally characterized 13 designs and obtained a monomeric protein with a weak fluorescence. We enhanced the fluorescence of this designed protein through two rounds of directed evolution and obtained designed monomeric smURFP (DMsmURFP), a bright, stable, and monomeric NIR FP with a molecular weight of 19.6 kDa. We determined the crystal structures of DMsmURFP both in the apo state and in complex with biliverdin, which confirmed the designed structure. The use of DMsmURFP in in vivo imaging of mammalian systems was demonstrated. The backbone design-based strategy used here can also be applied to monomerize other naturally multimeric proteins with intersubunit functional sites.
- 29 March 2024
- Published by American Chemical Society (ACS) in ACS Synthetic Biology
- Vol. 13 (4) , 1177-1190
Using Protein Design and Directed Evolution to Monomerize a Bright Near-Infrared Fluorescent Protein
The small ultrared fluorescent protein (smURFP) is a bright near-infrared (NIR) fluorescent protein (FP) that forms a dimer and binds its fluorescence chromophore, biliverdin, at its dimer interface. To engineer a monomeric NIR FP based on smURFP potentially more suitable for bioimaging, we employed protein design to extend the protein backbone with a new segment of two helices that shield the original dimer interface while covering the biliverdin binding pocket in place of the second chain in the original dimer. We experimentally characterized 13 designs and obtained a monomeric protein with a weak fluorescence. We enhanced the fluorescence of this designed protein through two rounds of directed evolution and obtained designed monomeric smURFP (DMsmURFP), a bright, stable, and monomeric NIR FP with a molecular weight of 19.6 kDa. We determined the crystal structures of DMsmURFP both in the apo state and in complex with biliverdin, which confirmed the designed structure. The use of DMsmURFP in in vivo imaging of mammalian systems was demonstrated. The backbone design-based strategy used here can also be applied to monomerize other naturally multimeric proteins with intersubunit functional sites.
De novo design of cavity-containing proteins with a backbone-centered neural network energy function