Zhenjiao Du
Zhenjiao Du has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Yonghui Li
Kansas State University
10 shared publications Xiaolong Guo
Department of Electrical and Computer Engineering
4 shared publicationsDoina Caragea
Department of Computer Science, Kansas State University
3 shared publications- Donghai Wang
Department of Biological and Agricultural Engineering, Kansas State University, Manhattan, KS 66506, USA
2 shared publications Kansas State University
2 shared publications- Nandan Kumar
Kansas State University
2 shared publications
Latest Publications
- 31 October 2026
- Published by Elsevier in Food Hydrocolloids
- Vol. 185
One-step preparation of drinkable yogurt via fucoidan-regulated casein gelation: Mechanism, stability and quality
- 31 October 2026
- Published by Elsevier in Food Hydrocolloids
- Vol. 185
One-step preparation of drinkable yogurt via fucoidan-regulated casein gelation: Mechanism, stability and quality
- Open Access
- 7 September 2026
- Published by Elsevier in The Innovation
- Vol. 7 (11)
AI can cheat scientists but never itself
- 7 September 2026
- Published by Elsevier in The Innovation
- Vol. 7 (11)
Open AccessAI can cheat scientists but never itself
- 9 June 2026
- Published by Taylor & Francis in Critical Reviews in Food Science and Nutrition
- Vol. 66 (30) , 1-24
Artificial intelligence in food allergen detection and prediction: advances, methodologies, and challenges
Food allergies affect over 220 million individuals worldwide and present increasing challenges due to complex food matrices and processing-induced protein modifications. Conventional detection methods, including immunoassays, PCR, and mass spectrometry, provide reliable analytical tools but are often limited by matrix interference, cross-reactivity, and labor-intensive workflows. Artificial intelligence (AI) has emerged as a complementary strategy, enabling high-throughput allergen prediction and enhanced analytical signal interpretation. This review examines recent advances in AI-driven allergen research across computational prediction and analytical detection. Machine learning (ML) and deep learning (DL) models achieve predictive accuracies exceeding 90% in sequence-based allergenicity assessment, outperforming traditional similarity-based methods. In analytical systems, AI-assisted spectroscopy and imaging enable rapid detection within seconds to minutes. Despite these advances, challenges remain in dataset bias, model interpretability, and cross-domain generalization. Future work should focus on explainable AI, standardized datasets, and external validation to support reliable and deployable allergen risk management systems. The integration of AI with spectroscopy, imaging, biosensing, and mass spectrometry is also highlighted.
- 9 June 2026
- Published by Taylor & Francis in Critical Reviews in Food Science and Nutrition
- Vol. 66 (30) , 1-24
Artificial intelligence in food allergen detection and prediction: advances, methodologies, and challenges
Food allergies affect over 220 million individuals worldwide and present increasing challenges due to complex food matrices and processing-induced protein modifications. Conventional detection methods, including immunoassays, PCR, and mass spectrometry, provide reliable analytical tools but are often limited by matrix interference, cross-reactivity, and labor-intensive workflows. Artificial intelligence (AI) has emerged as a complementary strategy, enabling high-throughput allergen prediction and enhanced analytical signal interpretation. This review examines recent advances in AI-driven allergen research across computational prediction and analytical detection. Machine learning (ML) and deep learning (DL) models achieve predictive accuracies exceeding 90% in sequence-based allergenicity assessment, outperforming traditional similarity-based methods. In analytical systems, AI-assisted spectroscopy and imaging enable rapid detection within seconds to minutes. Despite these advances, challenges remain in dataset bias, model interpretability, and cross-domain generalization. Future work should focus on explainable AI, standardized datasets, and external validation to support reliable and deployable allergen risk management systems. The integration of AI with spectroscopy, imaging, biosensing, and mass spectrometry is also highlighted.
- 25 April 2025
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry B
- Vol. 129 (18) , 4392-4409
Membrane Selectivity Mechanisms of the Antimicrobial Peptide Snakin-Z Against Prokaryotic and Eukaryotic Membrane Models
Snakin-Z, a novel cationic antimicrobial peptide (AMP) derived from Zizyphus jujuba fruits, exhibits broad-spectrum antimicrobial activity against bacteria and fungi. Importantly, it displays minimal hemolytic activity toward human red blood cells (RBCs). Elucidating the molecular basis of membrane selectivity of Snakin-Z is essential for its development as a novel antimicrobial agent. In this study, all-atom molecular dynamics (MD) simulations were employed to provide detailed molecular insights into the interactions between Snakin-Z and bacterial, fungal, and RBC membrane models. The simulations revealed a helical-coil conformation for Snakin-Z, with its amphipathic structure, polarity, and residues such as Arg, Lys, Ser, and Tyr playing crucial roles in mediating selective interactions with the microbial membrane models. Specifically, Arg28, Lys29, and Arg3 were identified as playing a crucial role in mediating membrane binding and stability. Snakin-Z was observed to be deeply embedded in the Candida albicans and Bacillus subtilis membrane models, followed by Escherichia coli and RBC membrane models. A considerable thinning and strong disordering of Candida albicans, Bacillus subtilis and Escherichia coli membranes acyl chains were observed. The presence of cholesterol in the RBC membrane contributes to its resistance to Snakin-Z-mediated disruption. This study presents the first comprehensive investigation of the selective mechanism underlying the antimicrobial activity of Snakin-Z against bacterial membrane models. Our findings provide insights into the antimicrobial properties of Snakin-Z at the molecular level, highlighting its significant potential for use in the food and biotechnology industries as a promising alternative to conventional antibiotics and preservatives.
- 25 April 2025
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry B
- Vol. 129 (18) , 4392-4409
Membrane Selectivity Mechanisms of the Antimicrobial Peptide Snakin-Z Against Prokaryotic and Eukaryotic Membrane Models
Snakin-Z, a novel cationic antimicrobial peptide (AMP) derived from Zizyphus jujuba fruits, exhibits broad-spectrum antimicrobial activity against bacteria and fungi. Importantly, it displays minimal hemolytic activity toward human red blood cells (RBCs). Elucidating the molecular basis of membrane selectivity of Snakin-Z is essential for its development as a novel antimicrobial agent. In this study, all-atom molecular dynamics (MD) simulations were employed to provide detailed molecular insights into the interactions between Snakin-Z and bacterial, fungal, and RBC membrane models. The simulations revealed a helical-coil conformation for Snakin-Z, with its amphipathic structure, polarity, and residues such as Arg, Lys, Ser, and Tyr playing crucial roles in mediating selective interactions with the microbial membrane models. Specifically, Arg28, Lys29, and Arg3 were identified as playing a crucial role in mediating membrane binding and stability. Snakin-Z was observed to be deeply embedded in the Candida albicans and Bacillus subtilis membrane models, followed by Escherichia coli and RBC membrane models. A considerable thinning and strong disordering of Candida albicans, Bacillus subtilis and Escherichia coli membranes acyl chains were observed. The presence of cholesterol in the RBC membrane contributes to its resistance to Snakin-Z-mediated disruption. This study presents the first comprehensive investigation of the selective mechanism underlying the antimicrobial activity of Snakin-Z against bacterial membrane models. Our findings provide insights into the antimicrobial properties of Snakin-Z at the molecular level, highlighting its significant potential for use in the food and biotechnology industries as a promising alternative to conventional antibiotics and preservatives.
- Open Access
PepBERT: Lightweight language models for bioactive peptide representation
Protein language models (pLMs) have been widely adopted for various protein and peptide-related downstream tasks and demonstrated promising performance. However, short peptides are significantly underrepresented in commonly used pLM training datasets. For example, only 2.8% of sequences in the UniProt Reference Cluster (UniRef) contain fewer than 50 residues, which potentially limits the effectiveness of pLMs for peptide-specific applications. Here, we present PepBERT, a lightweight and efficient peptide language model specifically designed for encoding peptide sequences. Two versions of the model—PepBERT-large (4.9 million parameters) and PepBERT-small (1.86 million parameters)—were pretrained from scratch using four custom peptide datasets and evaluated on nine peptide-related downstream prediction tasks. Both PepBERT models achieved performance superior to or comparable to the benchmark model, ESM-2 with 7.5 million parameters, on 8 out of 9 datasets. Overall, PepBERT provides a compact yet effective solution for generating high-quality peptide representations for downstream applications. By enabling more accurate representation and prediction of bioactive peptides, PepBERT can accelerate the discovery of food-derived bioactive peptides with health-promoting properties, supporting the development of sustainable functional foods and value-added utilization of food processing by-products. The datasets, source codes, pretrained models, and tutorials for the usage of PepBERT are available at https://github.com/dzjxzyd/PepBERT .
Open AccessPepBERT: Lightweight language models for bioactive peptide representation
Protein language models (pLMs) have been widely adopted for various protein and peptide-related downstream tasks and demonstrated promising performance. However, short peptides are significantly underrepresented in commonly used pLM training datasets. For example, only 2.8% of sequences in the UniProt Reference Cluster (UniRef) contain fewer than 50 residues, which potentially limits the effectiveness of pLMs for peptide-specific applications. Here, we present PepBERT, a lightweight and efficient peptide language model specifically designed for encoding peptide sequences. Two versions of the model—PepBERT-large (4.9 million parameters) and PepBERT-small (1.86 million parameters)—were pretrained from scratch using four custom peptide datasets and evaluated on nine peptide-related downstream prediction tasks. Both PepBERT models achieved performance superior to or comparable to the benchmark model, ESM-2 with 7.5 million parameters, on 8 out of 9 datasets. Overall, PepBERT provides a compact yet effective solution for generating high-quality peptide representations for downstream applications. By enabling more accurate representation and prediction of bioactive peptides, PepBERT can accelerate the discovery of food-derived bioactive peptides with health-promoting properties, supporting the development of sustainable functional foods and value-added utilization of food processing by-products. The datasets, source codes, pretrained models, and tutorials for the usage of PepBERT are available at https://github.com/dzjxzyd/PepBERT .
- 4 March 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 65 (6) , 2806-2817
FusionESP: Improved Enzyme–Substrate Pair Prediction by Fusing Protein and Chemical Knowledge
To reduce the cost of the experimental characterization of the potential substrates for enzymes, machine learning prediction models offer an alternative solution. Pretrained language models, as powerful approaches for protein and molecule representation, have been employed in the development of enzyme-substrate prediction models, achieving promising performance. In addition to continuing improvements in language models, effectively fusing encoders to handle multimodal prediction tasks is critical for further enhancing model performance by using available representation methods. Here, we present FusionESP, a multimodal architecture that integrates protein and chemistry language models with two independent projection heads and a contrastive learning strategy for predicting enzyme-substrate pairs. Our best model achieved state-of-the-art performance with an accuracy of 94.77% on independent test data and exhibited better generalization capacity while requiring fewer computational resources and training data, compared to previous studies of a fine-tuned encoder or employing more encoders. It also confirmed our hypothesis that embeddings of positive pairs are closer to each other in a high-dimension space, while negative pairs exhibit the opposite trend. Our ablation studies showed that the projection heads played a crucial role in performance enhancement, while the contrastive learning strategy further improved the projection heads' capacity in classification tasks. The proposed architecture is expected to be further applied to enhance performance in additional multimodality prediction tasks in biology. A user-friendly web server of FusionESP is established and freely accessible at https://rqkjkgpsyu.us-east-1.awsapprunner.com/.
- 4 March 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 65 (6) , 2806-2817
FusionESP: Improved Enzyme–Substrate Pair Prediction by Fusing Protein and Chemical Knowledge
To reduce the cost of the experimental characterization of the potential substrates for enzymes, machine learning prediction models offer an alternative solution. Pretrained language models, as powerful approaches for protein and molecule representation, have been employed in the development of enzyme-substrate prediction models, achieving promising performance. In addition to continuing improvements in language models, effectively fusing encoders to handle multimodal prediction tasks is critical for further enhancing model performance by using available representation methods. Here, we present FusionESP, a multimodal architecture that integrates protein and chemistry language models with two independent projection heads and a contrastive learning strategy for predicting enzyme-substrate pairs. Our best model achieved state-of-the-art performance with an accuracy of 94.77% on independent test data and exhibited better generalization capacity while requiring fewer computational resources and training data, compared to previous studies of a fine-tuned encoder or employing more encoders. It also confirmed our hypothesis that embeddings of positive pairs are closer to each other in a high-dimension space, while negative pairs exhibit the opposite trend. Our ablation studies showed that the projection heads played a crucial role in performance enhancement, while the contrastive learning strategy further improved the projection heads' capacity in classification tasks. The proposed architecture is expected to be further applied to enhance performance in additional multimodality prediction tasks in biology. A user-friendly web server of FusionESP is established and freely accessible at https://rqkjkgpsyu.us-east-1.awsapprunner.com/.
- 29 January 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 65 (3) , 1128-1139
pLM4CPPs: Protein Language Model-Based Predictor for Cell Penetrating Peptides
Cell-penetrating peptides (CPPs) are short peptides capable of penetrating cell membranes, making them valuable for drug delivery and intracellular targeting. Accurate prediction of CPPs can streamline experimental validation in the lab. This study aims to assess pretrained protein language models (pLMs) for their effectiveness in representing CPPs and develop a reliable model for CPP classification. We evaluated peptide embeddings generated from BEPLER, CPCProt, SeqVec, various ESM variants (ESM, ESM-2 with expanded feature set, ESM-1b, and ESM-1v), ProtT5-XL UniRef50, ProtT5-XL BFD, and ProtBERT. We developed pLM4CCPs, a novel deep learning architecture using convolutional neural networks (CNNs) as the classifier for binary classification of CPPs. pLM4CCPs demonstrated superior performance over existing state-of-the-art CPP prediction models, achieving improvements in accuracy (ACC) by 4.9-5.5%, Matthews correlation coefficient (MCC) by 9.3-10.2%, and sensitivity (Sn) by 14.1-19.6%. Among all the tested models, ESM-1280 and ProtT5-XL BFD demonstrated the highest overall performance on the kelm data set. ESM-1280 achieved an ACC of 0.896, an MCC of 0.796, a Sn of 0.844, and a specificity (Sp) of 0.978. ProtT5-XL BFD exhibited superior performance with an ACC of 0.901, an MCC of 0.802, an Sn of 0.885, and an Sp of 0.917. pLM4CCPs combine predictions from multiple models to provide a consensus on whether a given peptide sequence is classified as a CPP or non-CPP. This approach will enhance prediction reliability by leveraging the strengths of each individual model. A user-friendly web server for bioactivity predictions, along with data sets, is available at https://ry2acnp6ep.us-east-1.awsapprunner.com. The source code and protocol for adapting pLM4CPPs can be accessed on GitHub at https://github.com/drkumarnandan/pLM4CPPs. This platform aims to advance CPP prediction and peptide functionality modeling, aiding researchers in exploring peptide functionality effectively.
- 29 January 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 65 (3) , 1128-1139
pLM4CPPs: Protein Language Model-Based Predictor for Cell Penetrating Peptides
Cell-penetrating peptides (CPPs) are short peptides capable of penetrating cell membranes, making them valuable for drug delivery and intracellular targeting. Accurate prediction of CPPs can streamline experimental validation in the lab. This study aims to assess pretrained protein language models (pLMs) for their effectiveness in representing CPPs and develop a reliable model for CPP classification. We evaluated peptide embeddings generated from BEPLER, CPCProt, SeqVec, various ESM variants (ESM, ESM-2 with expanded feature set, ESM-1b, and ESM-1v), ProtT5-XL UniRef50, ProtT5-XL BFD, and ProtBERT. We developed pLM4CCPs, a novel deep learning architecture using convolutional neural networks (CNNs) as the classifier for binary classification of CPPs. pLM4CCPs demonstrated superior performance over existing state-of-the-art CPP prediction models, achieving improvements in accuracy (ACC) by 4.9-5.5%, Matthews correlation coefficient (MCC) by 9.3-10.2%, and sensitivity (Sn) by 14.1-19.6%. Among all the tested models, ESM-1280 and ProtT5-XL BFD demonstrated the highest overall performance on the kelm data set. ESM-1280 achieved an ACC of 0.896, an MCC of 0.796, a Sn of 0.844, and a specificity (Sp) of 0.978. ProtT5-XL BFD exhibited superior performance with an ACC of 0.901, an MCC of 0.802, an Sn of 0.885, and an Sp of 0.917. pLM4CCPs combine predictions from multiple models to provide a consensus on whether a given peptide sequence is classified as a CPP or non-CPP. This approach will enhance prediction reliability by leveraging the strengths of each individual model. A user-friendly web server for bioactivity predictions, along with data sets, is available at https://ry2acnp6ep.us-east-1.awsapprunner.com. The source code and protocol for adapting pLM4CPPs can be accessed on GitHub at https://github.com/drkumarnandan/pLM4CPPs. This platform aims to advance CPP prediction and peptide functionality modeling, aiding researchers in exploring peptide functionality effectively.
- Open Access
FusionESP: Improved enzyme-substrate pair prediction by fusing protein and chemical knowledge
To reduce the cost of experimental characterization of the potential substrates for enzymes, machine learning prediction model offers an alternative solution. Pretrained language models, as powerful approaches for protein and molecule representation, have been employed in the development of enzyme-substrate prediction models, achieving promising performance. In addition to continuing improvements in language models, effectively fusing encoders to handle multimodal prediction tasks is critical for further enhancing model performance using available representation methods. Here, we present FusionESP, a multimodal architecture that integrates protein and chemistry language models with a newly designed contrastive learning strategy for predicting enzyme-substrate pairs. Our best model achieved state-of-the-art performance with an accuracy of 94.77% on independent test data and exhibited better generalization capacity while requiring fewer computational resources and training data, compared to previous studies of finetuned encoder or employing more encoders. It also confirmed our hypothesis that embeddings of positive pairs are closer to each other in high-dimension space, while negative pairs exhibit the opposite trend. The proposed architecture is expected to be further applied to enhance performance in additional multimodality prediction tasks in biology. A user-friendly web server of FusionESP is established and freely accessible at https://rqkjkgpsyu.us-east-1.awsapprunner.com/ .
Open AccessFusionESP: Improved enzyme-substrate pair prediction by fusing protein and chemical knowledge
To reduce the cost of experimental characterization of the potential substrates for enzymes, machine learning prediction model offers an alternative solution. Pretrained language models, as powerful approaches for protein and molecule representation, have been employed in the development of enzyme-substrate prediction models, achieving promising performance. In addition to continuing improvements in language models, effectively fusing encoders to handle multimodal prediction tasks is critical for further enhancing model performance using available representation methods. Here, we present FusionESP, a multimodal architecture that integrates protein and chemistry language models with a newly designed contrastive learning strategy for predicting enzyme-substrate pairs. Our best model achieved state-of-the-art performance with an accuracy of 94.77% on independent test data and exhibited better generalization capacity while requiring fewer computational resources and training data, compared to previous studies of finetuned encoder or employing more encoders. It also confirmed our hypothesis that embeddings of positive pairs are closer to each other in high-dimension space, while negative pairs exhibit the opposite trend. The proposed architecture is expected to be further applied to enhance performance in additional multimodality prediction tasks in biology. A user-friendly web server of FusionESP is established and freely accessible at https://rqkjkgpsyu.us-east-1.awsapprunner.com/ .
- 19 December 2023
- Published by American Chemical Society (ACS) in Journal of Agricultural and Food Chemistry
- Vol. 72 (1) , 752-760
pLM4Alg: Protein Language Model-Based Predictors for Allergenic Proteins and Peptides
The rising prevalence of allergy demands efficient and accurate bioinformatic tools to expedite allergen identification and risk assessment while also reducing wet experiment expenses and time. Recently, pretrained protein language models (pLMs) have successfully predicted protein structure and function. However, to our best knowledge, they have not been used for predicting allergenic proteins/peptides. Therefore, this study aims to develop robust models for allergenic protein/peptide prediction using five pLMs of varying sizes and systematically assess their performance through fine-tuning with a convolutional neural network. The developed pLM4Alg models have achieved state-of-the-art performance with accuracy, Matthews correlation coefficient, and area under the curve scoring 93.4-95.1%, 0.869-0.902, and 0.981-0.990, respectively. Moreover, pLM4Alg is the first model capable of handling prediction tasks involving residue-missed sequences and sequences containing nonstandard amino acid residues. To facilitate easy access, a user-friendly web server (https://f6wxpfd3sh.us-east-1.awsapprunner.com) has been established. pLM4Alg is expected to become the leading machine learning-based prediction model for allergenic peptides and proteins. Its collaboration with other predictors holds great promise for accelerating allergy research.
- 19 December 2023
- Published by American Chemical Society (ACS) in Journal of Agricultural and Food Chemistry
- Vol. 72 (1) , 752-760
pLM4Alg: Protein Language Model-Based Predictors for Allergenic Proteins and Peptides
The rising prevalence of allergy demands efficient and accurate bioinformatic tools to expedite allergen identification and risk assessment while also reducing wet experiment expenses and time. Recently, pretrained protein language models (pLMs) have successfully predicted protein structure and function. However, to our best knowledge, they have not been used for predicting allergenic proteins/peptides. Therefore, this study aims to develop robust models for allergenic protein/peptide prediction using five pLMs of varying sizes and systematically assess their performance through fine-tuning with a convolutional neural network. The developed pLM4Alg models have achieved state-of-the-art performance with accuracy, Matthews correlation coefficient, and area under the curve scoring 93.4-95.1%, 0.869-0.902, and 0.981-0.990, respectively. Moreover, pLM4Alg is the first model capable of handling prediction tasks involving residue-missed sequences and sequences containing nonstandard amino acid residues. To facilitate easy access, a user-friendly web server (https://f6wxpfd3sh.us-east-1.awsapprunner.com) has been established. pLM4Alg is expected to become the leading machine learning-based prediction model for allergenic peptides and proteins. Its collaboration with other predictors holds great promise for accelerating allergy research.
- 8 November 2022
- Published by American Chemical Society (ACS) in ACS Food Science & Technology
- Vol. 2 (11) , 1781-1788
Computer-Aided Approaches for Screening Antioxidative Dipeptides and Application to Sorghum Proteins
- 8 November 2022
- Published by American Chemical Society (ACS) in ACS Food Science & Technology
- Vol. 2 (11) , 1781-1788
Computer-Aided Approaches for Screening Antioxidative Dipeptides and Application to Sorghum Proteins