Protein targeting, often guided by targeting peptides, is a critical biological process that directs proteins to their specific cellular destinations, ensuring proper cellular functionality and organization. Accurate classification and detection of targeting peptides are fundamental to understanding protein sorting mechanisms. This study introduces MULoc-Target, a novel deep-learning method designed to detect and classify targeting peptides in eukaryotic proteins. To support its development and evaluation, we curated a benchmark dataset comprising eight types of eukaryotic targeting peptides with manually curated annotations. Comprehensive evaluations on this dataset and external datasets from the literature demonstrate that MULoc-Target achieves state-of-the-art or competitive performance in detecting and classifying targeting peptides. Additionally, it enables the extraction of enriched motif patterns, offering valuable insights into their properties and the underlying targeting mechanisms. The identified motifs align closely with established biological features, further validating MULoc-Target's capabilities. A web server for MULoc-Target is integrated into our MULocDeep localization suite as a new toolkit, publicly accessible at https://mu-loc.org/MULoc-Target, and the inference code is available at https://github.com/yuexujiang/MULoc-Target.

We present a lightweight and efficient semi-supervised video object segmentation network based on the space-time memory framework. To some extent, our method solves the two difficulties encountered in traditional video object segmentation: one is that the single frame calculation time is too long, and the other is that the current frame’s segmentation should use more information from past frames. The algorithm uses a global context (GC) module to achieve high-performance, real-time segmentation. The GC module can effectively integrate multi-frame image information without increased memory and can process each frame in real time. Moreover, the prediction mask of the previous frame is helpful for the segmentation of the current frame, so we input it into a spatial constraint module (SCM), which constrains the areas of segments in the current frame. The SCM effectively alleviates mismatching of similar targets yet consumes few additional resources. We added a refinement module to the decoder to improve boundary segmentation. Our model achieves state-of-the-art results on various datasets, scoring 80.1% on YouTube-VOS 2018 and a $${\cal J}{\rm{\& }}{\cal F}$$ score of 78.0% on DAVIS 2017, while taking 0.05 s per frame on the DAVIS 2016 validation dataset.

Protein-peptide interactions play a crucial role in a wide range of biological processes, such as cellular regulation, immune responses, and signal transduction. Understanding the details of these interactions and predicting their complex structures is a vital key to peptide-based drug design. Predicting protein-peptide docking structure has experienced impressive scientific momentum over the past few years, which plays a crucial role in designing and developing peptide drugs. A wide range of computational algorithms has been developed for protein-peptide-docking predictions, but almost all of them require experimental protein structures, which are expensive and time-consuming. Benefiting from Alphafold2 and RoseTTAFold tools, most protein structures can be accurately deciphered based on protein sequences only. Formulating protein-peptide binding/docking as a protein complex folding problem, we can use Alphafold2 to generate a series of bounded protein-peptide conformations. We have designed a pipeline for predicting protein-peptide complexes and scoring the predicted models using AI-based tools. We benchmarked the pipeline by a set of non-redundant protein-peptide complex structures derived from the most recent-released complexes in PDB during the 2021 and 2022 years. Both template-based and template-free methods of Alphafold2 were specifically analyzed in protein-peptide complex structure predictions, and the strengths and weaknesses of each method were identified. We showed that the near-native complex structures were often not obtained by both of them and suggested that integrating the results of both template-based and template-free methods could be a good strategy. We compared the results with several protein-peptide tools, such as RoseTTAFold. We also evaluated several scoring schemes, including our in-house method based on a graph neural network, in ranking protein-peptide binding conformations.

MOTIVATION: As large amounts of biological data continue to be rapidly generated, a major focus of bioinformatics research has been aimed toward integrating these data to identify active pathways or modules under certain experimental conditions or phenotypes. Although biologically significant modules can often be detected globally by many existing methods, it is often hard to interpret or make use of the results toward pathway model generation and testing. RESULTS: To address this gap, we have developed the IMPRes algorithm, a new step-wise active pathway detection method using a dynamic programing approach. IMPRes takes advantage of the existing pathway interaction knowledge in Kyoto Encyclopedia of Genes and Genomes. Omics data are then used to assign penalties to genes, interactions and pathways. Finally, starting from one or multiple seed genes, a shortest path algorithm is applied to detect downstream pathways that best explain the gene expression data. Since dynamic programing enables the detection one step at a time, it is easy for researchers to trace the pathways, which may lead to more accurate drug design and more effective treatment strategies. The evaluation experiments conducted on three yeast datasets have shown that IMPRes can achieve competitive or better performance than other state-of-the-art methods. Furthermore, a case study on human lung cancer dataset was performed and we provided several insights on genes and mechanisms involved in lung cancer, which had not been discovered before. AVAILABILITY AND IMPLEMENTATION: IMPRes visualization tool is available via web server at http://digbio.missouri.edu/impres. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Zinc is an important cofactor for various biological functions in plants and animals, which are usually associated with proteins. Zinc also plays an important role in protein structures to which it binds. Hence, it is important to predict the Zinc binding sites in these proteins to better understand the structures and functions of these proteins. Most of the existing tools developed in this domain are structure-based predictors implementing Support Vector Machines on datasets that are more than a decade old. As there is little work done to explore the use of deep learning frameworks in this problem, we propose ZinCaps, a framework based on the capsule network for predicting zinc binding site using sequence-only information on more recently compiled datasets. ZinCaps outperforms previous tools. Its source codes is freely available for download at https://github.com/clemEssien/ActiveSitePrediction.

Ubiquitylation, a typical post-translational modification (PTM), plays an important role in signal transduction, apoptosis and cell proliferation. A ubiquitylation like PTM, sumoylation also may affect gene mapping, expression and genomic replication. Over the past two decades, machine learning has been widely employed in protein ubiquitylation and sumoylation site prediction tools. These existing tools require feature engineering, but failed to provide general interpretable features and probably underutilized the growing amount of data. This prompted us to propose a deep learning-based model that integrates multiple convolution and fully-connected layers of seven supervised learning sub-models to extract deep representations from protein sequences and physico-chemical properties (PCPs). Especially, we divided PCPs into 6 clusters and customized deep networks accordingly for handling the high correlations among one cluster. A stacking ensemble strategy was applied to combine these deep representations to make prediction. Furthermore, with the advantage of transfer learning, our deep learning model can work well on protein sumoylation site prediction as well after fine-tuning. On the high-quality annotated database Swiss-Prot, our model outperformed several well-known ubiquitylation and sumoylation site prediction tools. Our code is freely available at https://github.com/ruiwcoding/DeepUbiSumoPre.

Many new molecular mechanisms of genomics, pharmacogenomics, immunology, and other fields are reflected in pathway figures and need to be curated for various applications, especially in precision medicine. Current manual curation approaches are inadequate in keeping up with the pace of biomedical literature growth. Compared with textual representations, pathway figures in the biomedical literature often contain more direct representations of the mechanisms. However, no systematic method for curating pathway figures exists in publications due to date. Here, we propose a pathway curation pipeline, which integrates a deep learning model with an optical character recognition method and an image processing strategy to capture the locations, names, and interactions of pathway entities in the figure. Our pipeline was evaluated on the figures from PubMed publications. The results demonstrate that our model can effectively retrieve molecular entities and their interactions from pathway figures at a large scale. The proposed pipeline provides a complementary approach to text-mining in biological literature mining. In future work, we will combine our method with text-mining tools to enrich extracted information and reconstruct pathway mechanisms more precisely.

MOTIVATION: Computational methods for protein post-translational modification (PTM) site prediction provide a useful approach for studying protein functions. The prediction accuracy of the existing methods has significant room for improvement. A recent deep-learning architecture, Capsule Network (CapsNet), which can characterize the internal hierarchical representation of input data, presents a great opportunity to solve this problem, especially using small training data. RESULTS: We proposed a CapsNet for predicting protein PTM sites, including phosphorylation, N-linked glycosylation, N6-acetyllysine, methyl-arginine, S-palmitoyl-cysteine, pyrrolidone-carboxylic-acid and SUMOylation sites. The CapsNet outperformed the baseline convolutional neural network architecture MusiteDeep and other well-known tools in most cases and provided promising results for practical use, especially in learning from small training data. The capsule length also gives an accurate estimate for the confidence of the PTM prediction. We further demonstrated that the internal capsule features could be trained as a motif detector of phosphorylation sites when no kinase-specific phosphorylation labels were provided. In addition, CapsNet generates robust representations that have strong discriminant power in distinguishing kinase substrates from different kinase families. Our study sheds some light on the recognition mechanism of PTMs and applications of CapsNet on other bioinformatic problems. AVAILABILITY AND IMPLEMENTATION: The codes are free to download from https://github.com/duolinwang/CapsNet_PTM. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

A central goal of systems biology is to uncover the underlying functional architecture of the cell and study its mechanisms. To this end, large amounts of omics data are being rapidly generated, and a focus of bioinformatics research has been towards integrating these data to identify active pathways or modules under certain conditions. Many bioinformatics algorithms include optimization methods, statistical methods, and methods using interaction network topology attributes have been applied for this. Although biologically significant modules can often be detected globally by these methods, it is hard to interpret or make use of the results towards in silico hypothesis generation and testing. We propose a step-wise active pathway detection method (IMPRes) using a dynamic programming approach. First, we take advantage of the existing pathway interaction knowledge in KEGG to build a background network, and then starting from one or multiple receptors of a certain perturbation, we use transcriptomics data collected under these conditions to detect paths that best explain the variations of genes downstream. More other omics data will be integrated in the future. Since dynamic programming enables the detection one step a time, it is easy for biomedical researchers to trace the pathway and finally lead to more accurate drug design and more effective treatment strategies. Additionally, by adding protein-protein interactions in our method, the hypotheses that we generate do not merely utilize existing knowledge, but have potential to discover new knowledge. We have evaluated our method on a dataset of cell wall stress in yeast. The path we found highly agrees with the Cell Wall Integrity (CWI) pathway, which is the main signaling pathway involved in the regulation of cell wall stress responses. We have also compared with other methods on a yeast high osmolality stress dataset and achieved an overall better performance than some other methods. More experiments have been done on human cancer datasets and mouse datasets. Finally, the IMPRes web server is established to offer a simple interface for applying IMPRes. Users can upload their own data and obtain an interactive visualization of the resulting pathway map. Users can further filter or highlight interactions according to pathway information or relation types. All genes in the pathway map are listed with detailed annotations. The IMPRes web server is available at http://gene.rnet.missouri.edu/soykb_dev/IMPRes/.

Genomic selection (GS) is a marker-assisted selection approach to enhance quantitative traits in breeding population in which whole genome single-nucleotide polymorphisms (SNPs) markers can be used to predict breeding values (BV). GS has been proved to increase breeding efficiency in both plant and animal breeding, such as dairy cattle, pig, rice, soybean and loblolly pine. Here, we propose a deep-learning model using convolutional neural network (CNN) to predict genomic estimated breeding value (GEBV) and also to investigate neighboring SNP effects within linkage disequilibrium. We have applied our models on two datasets: 1) grain yield (YLD) trait on Glycine Max (soybean) nested association mapping (NAM) dataset and 2) stem height (HT) trait on a Pinus taeda (loblolly pine) dataset. The SoyNAM population contains 4,313 SNPs from 5,139 individuals with a trait heritability of 0.345 and the Loblolly Pine population contains 4,853 SNPs from 861 individuals with a trait heritability of 0.31. Our deep-learning model was tested with a 10-fold cross-validation, run in parallel on graphic processing units (GPUs). Our model prediction accuracy, which was calculated by Pearson Correlation between GEBV and observed values, outperforms traditional statistical RR-BLUP, Bayesian LASSO and BayesA models. The results indicate that deep-learning model is efficient in accurately computing breeding values and simultaneously studying nearby SNP effects from CNNs. It also indicates powerful potential in interpreting phenotype-genotype associations over the entire genome.

Amidation plays an important role in a variety of pathological processes and serious diseases like neural dysfunction and hypertension. However, identification of protein amidation sites through traditional experimental methods is time consuming and expensive. Existing computational methods for predicting amidation sites are based on feature extraction, which may result in incomplete or biased features. To address these issues, we proposed a novel deep-learning predictor for amidation sites. Deep learning as the cutting-edge machine learning method has the ability to automatically discover complex representations of amidation patterns from the raw protein sequences, and hence it provides a powerful tool for improvement of amidation site prediction.

Computational methods for post-translational modification (PTM) site prediction play important roles in protein function studies and experimental design. Most existing methods are based on feature extraction, which may result in incomplete or biased features. Deep learning as the cutting-edge machine learning method has the ability to automatically discover complex representations of PTM patterns from the raw sequences, and hence it provides a powerful tool for improvement of post-translational modification site prediction. In our previous work, we proposed MusiteDeep, the first deep-learning framework for predicting the phosphorylation, one of the well-studied PTMs. The previous MusiteDeep takes protein raw sequence as the input and uses convolutional neural networks with a two-dimensional attention mechanism. It achieved over a 50% improvement in the area under the precision-recall curve in general phosphorylation site prediction and obtains competitive results in kinase-specific prediction compared to other well-known tools on the benchmark data. In this work, we extended to explore more types of PTMs, including acetylation, methylation and glycosylation. New deep-learning architecture and new training strategy are proposed for better performance. Web server for more PTM site predictions and complex motif visualization will be developed in the future.

MOTIVATION: Computational methods for phosphorylation site prediction play important roles in protein function studies and experimental design. Most existing methods are based on feature extraction, which may result in incomplete or biased features. Deep learning as the cutting-edge machine learning method has the ability to automatically discover complex representations of phosphorylation patterns from the raw sequences, and hence it provides a powerful tool for improvement of phosphorylation site prediction. RESULTS: We present MusiteDeep, the first deep-learning framework for predicting general and kinase-specific phosphorylation sites. MusiteDeep takes raw sequence data as input and uses convolutional neural networks with a novel two-dimensional attention mechanism. It achieves over a 50% relative improvement in the area under the precision-recall curve in general phosphorylation site prediction and obtains competitive results in kinase-specific prediction compared to other well-known tools on the benchmark data. AVAILABILITY AND IMPLEMENTATION: MusiteDeep is provided as an open-source tool available at https://github.com/duolinwang/MusiteDeep. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.