BERT-Kgly: A Bidirectional Encoder Representations From Transformers (BERT)-Based Model for Predicting Lysine Glycation Site for Homo sapiens

Glypred: Lysine Glycation Site Prediction via CCU-LightGBM-BiLSTM Framework with Multi-Head Attention Mechanism

Glycation, a type of posttranslational modification, preferentially occurs on lysine and arginine residues, impairing protein functionality and altering characteristics. This process is linked to diseases such as Alzheimer's, diabetes, and atherosclerosis. Traditional wet lab experiments are time-consuming, whereas machine learning has significantly streamlined the prediction of protein glycation sites. Despite promising results, challenges remain, including data imbalance, feature redundancy, and suboptimal classifier performance. This research introduces Glypred, a lysine glycation site prediction model combining ClusterCentroids Undersampling (CCU), LightGBM, and bidirectional long short-term memory network (BiLSTM) methodologies, with an additional multihead attention mechanism integrated into the BiLSTM. To achieve this, the study undertakes several key steps: selecting diverse feature types to capture comprehensive protein information, employing a cluster-based undersampling strategy to balance the data set, using LightGBM for feature selection to enhance model performance, and implementing a bidirectional LSTM network for accurate classification. Together, these approaches ensure that Glypred effectively identifies glycation sites with high accuracy and robustness. For feature encoding, five distinct feature types─AAC, KMER, DR, PWAA, and EBGW─were selected to capture a broad spectrum of protein sequence and biological information. These encoded features were integrated and validated to ensure comprehensive protein information acquisition. To address the issue of highly imbalanced positive and negative samples, various undersampling algorithms, including random undersampling, NearMiss, edited nearest neighbor rule, and CCU, were evaluated. CCU was ultimately chosen to remove redundant nonglycated training data, establishing a balanced data set that enhances the model's accuracy and robustness. For feature selection, the LightGBM ensemble learning algorithm was employed to reduce feature dimensionality by identifying the most significant features. This approach accelerates model training, enhances generalization capabilities, and ensures good transferability of the model. Finally, a bidirectional long short-term memory network was used as the classifier, with a network structure designed to capture glycation modification site features from both forward and backward directions. To prevent overfitting, appropriate regularization parameters and dropout rates were introduced, achieving efficient classification. Experimental results show that Glypred achieved optimal performance. This model provides new insights for bioinformatics and encourages the application of similar strategies in other fields. A lysine glycation site prediction software tool was also developed using the PyQt5 library, offering researchers an auxiliary screening tool to reduce workload and improve efficiency. The software and data sets are available on GitHub: https://github.com/ZBYnb/Glypred.

Protein feature engineering framework for AMPylation site prediction

AMPylation is a biologically significant yet understudied post-translational modification where an adenosine monophosphate (AMP) group is added to Tyrosine and Threonine residues primarily. While recent work has illuminated the prevalence and functional impacts of AMPylation, experimental identification of AMPylation sites remains challenging. Computational prediction techniques provide a faster alternative approach. The predictive performance of machine learning models is highly dependent on the features used to represent the raw amino acid sequences. In this work, we introduce a novel feature extraction pipeline to encode the key properties relevant to AMPylation site prediction. We utilize a recently published dataset of curated AMPylation sites to develop our feature generation framework. We demonstrate the utility of our extracted features by training various machine learning classifiers, on various numerical representations of the raw sequences extracted with the help of our framework. Tenfold cross-validation is used to evaluate the model's capability to distinguish between AMPylated and non-AMPylated sites. The top-performing set of features extracted achieved MCC score of 0.58, Accuracy of 0.8, AUC-ROC of 0.85 and F1 score of 0.73. Further, we elucidate the behaviour of the model on the set of features consisting of monogram and bigram counts for various representations using SHapley Additive exPlanations.

SumoPred-PLM: human SUMOylation and SUMO2/3 sites Prediction using Pre-trained Protein Language Model

SUMOylation is an essential post-translational modification system with the ability to regulate nearly all aspects of cellular physiology. Three major paralogues SUMO1, SUMO2 and SUMO3 form a covalent bond between the small ubiquitin-like modifier with lysine residues at consensus sites in protein substrates. Biochemical studies continue to identify unique biological functions for protein targets conjugated to SUMO1 versus the highly homologous SUMO2 and SUMO3 paralogues. Yet, the field has failed to harness contemporary AI approaches including pre-trained protein language models to fully expand and/or recognize the SUMOylated proteome. Herein, we present a novel, deep learning-based approach called SumoPred-PLM for human SUMOylation prediction with sensitivity, specificity, Matthew's correlation coefficient, and accuracy of 74.64%, 73.36%, 0.48% and 74.00%, respectively, on the CPLM 4.0 independent test dataset. In addition, this novel platform uses contextualized embeddings obtained from a pre-trained protein language model, ProtT5-XL-UniRef50 to identify SUMO2/3-specific conjugation sites. The results demonstrate that SumoPred-PLM is a powerful and unique computational tool to predict SUMOylation sites in proteins and accelerate discovery.

Substrate Prediction for RiPP Biosynthetic Enzymes via Masked Language Modeling and Transfer Learning

Ribosomally synthesized and post-translationally modified peptide (RiPP) biosynthetic enzymes often exhibit promiscuous substrate preferences that cannot be reduced to simple rules. Large language models are promising tools for predicting such peptide fitness landscapes. However, state-of-the-art protein language models are trained on relatively few peptide sequences. A previous study comprehensively profiled the peptide substrate preferences of LazBF (a two-component serine dehydratase) and LazDEF (a three-component azole synthetase) from the lactazole biosynthetic pathway. We demonstrated that masked language modeling of LazBF substrate preferences produced language model embeddings that improved downstream classification models of both LazBF and LazDEF substrates. Similarly, masked language modelling of LazDEF substrate preferences produced embeddings that improved the performance of classification models of both LazBF and LazDEF substrates. Our results suggest that the models learned functional forms that are transferable between distinct enzymatic transformations that act within the same biosynthetic pathway. Our transfer learning method improved performance and data efficiency in data-scarce scenarios. We then fine-tuned models on each data set and showed that the fine-tuned models provided interpretable insight that we anticipate will facilitate the design of substrate libraries that are compatible with desired RiPP biosynthetic pathways.

iGly-IDN: Identifying Lysine Glycation Sites in Proteins Based on Improved DenseNet

Lysine glycation is one of the most significant protein post-translational modifications, which changes the properties of the proteins and causes them to be dysfunctional. Accurately identifying glycation sites helps to understand the biological function and potential mechanism of glycation in disease treatments. Nonetheless, the experimental methods are ordinarily inefficient and costly, so effective computational methods need to be developed. In this study, we proposed the new model called iGly-IDN based on the improved densely connected convolutional networks (DenseNet). First, one hot encoding was adopted to obtain the original feature maps. Afterward, the improved DenseNet was adopted to capture feature information with the importance degrees during the feature learning. According to the experimental results, Acc reaches 66%, and Mathews correlation coefficient reaches 0.33 on the independent testing data set, which indicates that the iGly-IDN can provide more effective glycation site identification than the current predictors.

Prediction of linear B-cell epitopes based on protein sequence features and BERT embeddings

Linear B-cell epitopes (BCEs) play a key role in the development of peptide vaccines and immunodiagnostic reagents. Therefore, the accurate identification of linear BCEs is of great importance in the prevention of infectious diseases and the diagnosis of related diseases. The experimental methods used to identify BCEs are both expensive and time-consuming and they do not meet the demand for identification of large-scale protein sequence data. As a result, there is a need to develop an efficient and accurate computational method to rapidly identify linear BCE sequences. In this work, we developed the new linear BCE prediction method LBCE-BERT. This method is based on peptide chain sequence information and natural language model BERT embedding information, using an XGBoost classifier. The models were trained on three benchmark datasets. The model was training on three benchmark datasets for hyperparameter selection and was subsequently evaluated on several test datasets. The result indicate that our proposed method outperforms others in terms of AUROC and accuracy. The LBCE-BERT model is publicly available at: https://github.com/Lfang111/LBCE-BERT .

Leveraging transformers-based language models in proteome bioinformatics

In recent years, the rapid growth of biological data has increased interest in using bioinformatics to analyze and interpret this data. Proteomics, which studies the structure, function, and interactions of proteins, is a crucial area of bioinformatics. Using natural language processing (NLP) techniques in proteomics is an emerging field that combines machine learning and text mining to analyze biological data. Recently, transformer-based NLP models have gained significant attention for their ability to process variable-length input sequences in parallel, using self-attention mechanisms to capture long-range dependencies. In this review paper, we discuss the recent advancements in transformer-based NLP models in proteome bioinformatics and examine their advantages, limitations, and potential applications to improve the accuracy and efficiency of various tasks. Additionally, we highlight the challenges and future directions of using these models in proteome bioinformatics research. Overall, this review provides valuable insights into the potential of transformer-based NLP models to revolutionize proteome bioinformatics.

MSTL-Kace: Prediction of Prokaryotic Lysine Acetylation Sites Based on Multistage Transfer Learning Strategy

As one of the most important post-translational modifications (PTM), lysine acetylation (Kace) plays an important role in various biological activities. Traditional experimental methods for identifying Kace sites are inefficient and expensive. Instead, several machine learning methods have been developed for Kace site prediction, and hand-crafted features have been used to encode the protein sequences. However, there are still two challenges: the complex biological information may be under-represented by these manmade features and the small sample issue of some species needs to be addressed. We propose a novel model, MSTL-Kace, which was developed based on transfer learning strategy with pretrained bidirectional encoder representations from transformers (BERT) model. In this model, the high-level embeddings were extracted from species-specific BERT models, and a two-stage fine-tuning strategy was used to deal with small sample issue. Specifically, a domain-specific BERT model was pretrained using all of the sequences in our data sets, which was then fine-tuned, or two-stage fine-tuned based on the training data set of each species to obtain the species-specific BERT models. Afterward, the embeddings of residues were extracted from the fine-tuned model and fed to the different downstream learning algorithms. After comparison, the best model for the six prokaryotic species was built by using a random forest. The results for the independent test sets show that our model outperforms the state-of-the-art methods on all six species. The source codes and data for MSTL-Kace are available at https://github.com/leo97king/MSTL-Kace.

NeuroPpred-SVM: A New Model for Predicting Neuropeptides Based on Embeddings of BERT

Neuropeptides play pivotal roles in different physiological processes and are related to different kinds of diseases. Identification of neuropeptides is of great benefit for studying the mechanism of these physiological processes and the treatment of neurological disorders. Several state-of-the-art neuropeptide predictors have been developed by using a two-layer stacking ensemble algorithm. Although the two-layer stacking ensemble algorithm can improve the feature representability, these models are complex, which are not as efficient as the models based on one classifier. In this study, we proposed a new model, NeuroPpred-SVM, to predict neuropeptides based on the embeddings of Bidirectional Encoder Representations from Transformers and other sequential features by using a support vector machine (SVM). The experimental results indicate that our model achieved a cross-validation area under the receiver operating characteristic (AUROC) curve of 0.969 on the training data set and an AUROC of 0.966 on the independent test set. By comparing our model with the other four state-of-the-art models including NeuroPIpred, PredNeuroP, NeuroPpred-Fuse, and NeuroPpred-FRL on the independent test set, our model achieved the highest AUROC, Matthews correlation coefficient, accuracy, and specificity, which indicate that our model outperforms the existing models. We believed that NeuroPpred-SVM could be a useful tool for identifying neuropeptides with high accuracy and low cost. The data sets and Python code are available at https://github.com/liuyf-a/NeuroPpred-SVM.

ProtTrans-Glutar: Incorporating Features From Pre-trained Transformer-Based Models for Predicting Glutarylation Sites

Lysine glutarylation is a post-translational modification (PTM) that plays a regulatory role in various physiological and biological processes. Identifying glutarylated peptides using proteomic techniques is expensive and time-consuming. Therefore, developing computational models and predictors can prove useful for rapid identification of glutarylation. In this study, we propose a model called ProtTrans-Glutar to classify a protein sequence into positive or negative glutarylation site by combining traditional sequence-based features with features derived from a pre-trained transformer-based protein model. The features of the model were constructed by combining several feature sets, namely the distribution feature (from composition/transition/distribution encoding), enhanced amino acid composition (EAAC), and features derived from the ProtT5-XL-UniRef50 model. Combined with random under-sampling and XGBoost classification method, our model obtained recall, specificity, and AUC scores of 0.7864, 0.6286, and 0.7075 respectively on an independent test set. The recall and AUC scores were notably higher than those of the previous glutarylation prediction models using the same dataset. This high recall score suggests that our method has the potential to identify new glutarylation sites and facilitate further research on the glutarylation process.