PlantNh-Kcr: a deep learning model for predicting non-histone crotonylation sites in plants.

BACKGROUND: Lysine crotonylation (Kcr) is a crucial protein post-translational modification found in histone and non-histone proteins. It plays a pivotal role in regulating diverse biological processes in both animals and plants, including gene transcription and replication, cell metabolism and differentiation, as well as photosynthesis. Despite the significance of Kcr, detection of Kcr sites through biological experiments is often time-consuming, expensive, and only a fraction of crotonylated peptides can be identified. This reality highlights the need for efficient and rapid prediction of Kcr sites through computational methods. Currently, several machine learning models exist for predicting Kcr sites in humans, yet models tailored for plants are rare. Furthermore, no downloadable Kcr site predictors or datasets have been developed specifically for plants. To address this gap, it is imperative to integrate existing Kcr sites detected in plant experiments and establish a dedicated computational model for plants. RESULTS: Most plant Kcr sites are located on non-histones. In this study, we collected non-histone Kcr sites from five plants, including wheat, tabacum, rice, peanut, and papaya. We then conducted a comprehensive analysis of the amino acid distribution surrounding these sites. To develop a predictive model for plant non-histone Kcr sites, we combined a convolutional neural network (CNN), a bidirectional long short-term memory network (BiLSTM), and attention mechanism to build a deep learning model called PlantNh-Kcr. On both five-fold cross-validation and independent tests, PlantNh-Kcr outperformed multiple conventional machine learning models and other deep learning models. Furthermore, we conducted an analysis of species-specific effect on the PlantNh-Kcr model and found that a general model trained using data from multiple species outperforms species-specific models. CONCLUSION: PlantNh-Kcr represents a valuable tool for predicting plant non-histone Kcr sites. We expect that this model will aid in addressing key challenges and tasks in the study of plant crotonylation sites.

Improving thermostability of Bacillus amyloliquefaciens alpha-amylase by multipoint mutations.

The medium-temperature alpha-amylase of Bacillus amyloliquefaciens is widely used in the food and washing process. Enhancing the thermostability of alpha-amylases and investigating the mechanism of stability are important for enzyme industry development. The optimal temperature and pH of the wild-type BAA and mutant MuBAA (D28E/V118A/S187D/K370 N) were all 60 °C and 6.0, respectively. The mutant MuBAA showed better thermostability at 50 °C and 60 °C, with a specific activity of 206.61 U/mg, which was 99.1% greater than that of the wild-type. By analyzing predicted structures, the improving thermostability of the mutant MuBAA was mainly related to enhanced stabilization of a loop region in domain B via more calcium-binding sites and intramolecular interactions around Asp187. Furthermore, additional intramolecular interactions around sites 28 and 370 in domain A were also beneficial for improving thermostability. Additionally, the decrease of steric hindrance at the active cavity increased the specific activity of the mutant MuBAA. Improving the thermostability of BAA has theoretical reference values for the modification of alpha-amylases.

Computational identification of human ubiquitination sites using convolutional and recurrent neural networks.

Ubiquitination is a very important protein post-translational modification in humans, which is closely related to many human diseases such as cancers. Although some methods have been elegantly proposed to predict human ubiquitination sites, the accuracy of these methods is generally not very satisfactory. In order to improve the prediction accuracy of human ubiquitination sites, we propose a new ensemble method HUbipPred, which takes the binary encoding and physicochemical properties of amino acids as training features, and integrates two intensively trained convolutional neural networks and two recurrent neural networks to build the model. Finally, HUbiPred achieves AUC values of 0.852 and 0.844 in five-fold cross-validation and independent tests, respectively, which greatly improves the prediction accuracy compared to previous predictors. We also analyze the physicochemical properties of amino acids around ubiquitination sites, study the important roles of architectures (i.e., convolution, long short-term memory (LSTM) and fully connected hidden layers) in the networks for prediction performance, and also predict potential ubiquitination sites in humans using HUbiPred. The training and test datasets, predicted human ubiquitination sites, and source codes of HUbiPred are publicly available at https://github.com/amituofo-xf/HUbiPred.

Computational identification of ubiquitination sites in Arabidopsis thaliana using convolutional neural networks.

KEY MESSAGE: We developed two CNNs for predicting ubiquitination sites in Arabidopsis thaliana, demonstrated their competitive performance, analyzed amino acid physicochemical properties and the CNN structures, and predicted ubiquitination sites in Arabidopsis. As an important posttranslational protein modification, ubiquitination plays critical roles in plant physiology, including plant growth and development, biotic and abiotic stress, metabolism, and so on. A lot of ubiquitination site prediction models have been developed for human, mouse and yeast. However, there are few models to predict ubiquitination sites for the plant Arabidopsis thaliana. Based on this context, we proposed two convolutional neural network (CNN) based models for predicting ubiquitination sites in A. thaliana. The two models reach AUC (area under the ROC curve) values of 0.924 and 0.913 respectively in five-fold cross-validation, and 0.921 and 0.914 respectively in independent test, which outperform other models and demonstrate the competitive edge of them. We in-depth analyze the amino acid physicochemical properties in the neighboring sequence regions of the ubiquitination sites, and study the influence of the CNN structure to the prediction performance. Potential ubiquitination sites in the global Arbidopsis proteome are predicted using the two CNN models. To facilitate the community, the source code, training and test dataset, predicted ubiquitination sites in the Arbidopsis proteome are available at GitHub ( http://github.com/nongdaxiaofeng/CNNAthUbi ) for interest users.

DDAPRED: a computational method for predicting drug repositioning using regularized logistic matrix factorization.

Due to rising development costs and stagnant product outputs of traditional drug discovery methods, drug repositioning, which discovers new indications for existing drugs, has attracted increasing interest. Computational drug repositioning can integrate prioritization information and accelerate time lines even further. However, most existing methods for predicting drug repositioning have low precisions. The present article proposed a new method named DDAPRED (https://github.com/nongdaxiaofeng/DDAPRED) for drug repositioning prediction. The method integrated multiple sources of drug similarity and disease similarity information, and it used the regularized logistic matrix decomposition method to significantly improve the prediction performance. In 5-fold cross-validation, the areas under the receiver operating characteristic curve (AUROC) and the precision-recall curve (AUPRC) of DDAPRED reached 0.932 and 0.438, respectively, exceeding other methods. The present study also analyzed the parameters influencing the model performance and the effect of different drug similarity information in-depth, and it verified the treatment relationship of the top 50 predictions with unknown relationships in the training set, further demonstrating the practicability of our method.

Expression and characterization of a chitinase from Serratia marcescens.

A chitinase gene from Serratia marcescens was cloned and expressed in Escherichia coli BL21(DE3) and the properties of recombinant chitinase rCHI-2 were characterized. The optimum catalytic pH of rCHI-2 was 6.0. It was stable in the pH range of 6.0-9.0 and could maintain more than 90% of its relative enzyme activity after incubation at 37 °C for 1 h. The optimum catalytic temperature of the enzyme was 55 °C and 85% of enzyme activity was remained after incubation at 45 °C for 1 h. The activation energy of the thermal inactivation of the enzyme was 10.9 kJ/mol and the Michaelis-Menten constant was 3.2 g/L. The purified rCHI-2 was found to be highly stable at 45 °C with half-life (t1/2) of 289 min and thermodynamic parameters ΔH*, ΔG* and ΔS* revealed high affinity of rCHI-2 for chitin. Hg2+ was found to be able to inhibit the enzyme activity reversibly, while IC50 and inhibition constant of Hg2+ on the enzyme were 34.8 µmol/L and 44.6 µmol/L, respectively. Moreover, rCHI-2 could specifically hydrolyze colloidal chitin into GlcNAc2 as the major product.

Prediction of zinc-binding sites using multiple sequence profiles and machine learning methods.

The zinc (Zn2+) cofactor has been proven to be involved in numerous biological mechanisms and the zinc-binding site is recognized as one of the most important post-translation modifications in proteins. Therefore, accurate knowledge of zinc ions in protein structures can provide potential clues for elucidation of protein folding and functions. However, determining zinc-binding residues by experimental means is usually lab-intensive and associated with high cost in most cases. In this context, the development of computational tools for identifying zinc-binding sites is highly desired, especially in the current post-genomic era. In this work, we developed a novel zinc-binding site prediction method by combining several intensively-trained machine learning models. To establish an accurate and generative method, we downloaded all zinc-binding proteins from the Protein Data Bank and prepared a non-redundant dataset. Meanwhile, a well-prepared dataset by other groups was also used. Then, effective and complementary features were extracted from sequences and three-dimensional structures of these proteins. Moreover, several well-designed machine learning models were intensively trained to construct accurate models. To assess the performance, the obtained predictors were stringently benchmarked using the diverse zinc-binding sites. Furthermore, several state-of-the-art in silico methods developed specifically for zinc-binding sites were also evaluated and compared. The results confirmed that our method is very competitive in real world applications and could become a complementary tool to wet lab experiments. To facilitate research in the community, a web server and stand-alone program implementing our method were constructed and are publicly available at . The downloadable program of our method can be easily used for the high-throughput screening of potential zinc-binding sites across proteomes.

Mutagenesis on the surface of a ß-agarase from Vibrio sp. ZC-1 increased its thermo-stability.

The recombinant rAgaZC-1 was a family GH50 ß-agarase from Vibrio sp. ZC-1 (CICC 24670). In this paper, the mutant D622G (i.e., mutate the aspartic acid at position 622 to glycine) had better thermo-stability than rAgaZC-1, showing 1.5℃ higher T5010 (the temperature at which the half-time is 10 min) and 4-folds of half-time at 41℃, while they had almost same optimum temperature (38.5℃), optimum pH (pH6.0) and catalytic efficiency. Thermal deactivation kinetical analysis showed that D622G had higher activation energy for deactivation, enthalpy and Gibbs free energy than rAgaZC-1, indicating that more energy is required by D622G for deactivation. Substrate can protect agarase against thermal inactivation, especially D622G. Hence the yield of agarose hydrolysis catalyzed by D622G was higher than that by rAgaZC-1. The models of D622G and rAgaZC-1 predicted by homology modeling were compared to find that it is the improved distribution of surface electrostatic potential, great symmetric positive potential and more hydrophobic interactions of D622G that enhance the thermo-stability.

A New Extremely Halophilic, Calcium-Independent and Surfactant-Resistant Alpha-Amylase from Alkalibacterium sp. SL3.

A new α-amylase-encoding gene (amySL3) of glycoside hydrolase (GH) family 13 was identified in soda lake isolate Alkalibacterium sp. SL3. The deduced AmySL3 shares high identities (82-98%) with putative α-amylases from the genus Alkalibacterium, but has low identities (<53%) with functionally characterized counterparts. amySL3 was successfully expressed in Escherichia coli, and the recombinant enzyme (rAmySL3) was purified to electrophoretic homogeneity. The optimal temperature and pH of the activity of the purified rAmySL3 were determined to be 45°C and pH 7.5, respectively. rAmySL3 was found to be extremely halophilic, showing maximal enzyme activity at a nearly saturated concentration of NaCl. Its thermostability was greatly enhanced in the presence of 4 M NaCl, and it was highly stable in 5 M NaCl. Moreover, the enzyme did not require calcium ions for activity, and was strongly resistant to a range of surfactants and hydrophobic organic solvents. The major hydrolysis products of rAmySL3 from soluble starch were maltobiose and maltotriose. The high ratio of acidic amino acids and highly negative electrostatic potential surface might account for the halophilic nature of AmySL3. The extremely halophilic, calcium-independent, and surfactant-resistant properties make AmySL3 a promising candidate enzyme for both basic research and industrial applications.

Truncation of κ­carrageenase for higher κ­carrageenan oligosaccharides yield with improved enzymatic characteristics.

Carrageenase is useful for preparation of carrageenan oligosaccharides, which have significant bioactivity. We expressed a κ­carrageenase gene from Zobellia sp. ZL-4 in full-length (κ-ZL-4) or after truncation of the carbohydrate binding module and the Type-IX secretion module (κ-ZL-4-GH16). κ-ZL-4-GH16 showed a specific activity (134.22 U/mg) 1.93 times higher than that of κ-ZL-4, and its thermal and pH stability also increased. The best activity of κ-ZL-4-GH16 was presented at pH 3.0-6.0, which was lower than the optimal pH of reported κ-carrageenases. The enzyme-substrate affinity of κ-ZL-4-GH16 was higher than that of κ-ZL-4, demonstrated by its lower Michaelis-Menten constant (0.704 mg/mL at pH 6.0). Importantly, κ-ZL-4-GH16 released 10-fold more κ-carrageenan disaccharides than κ-ZL-4. The κ-carrageenan tetrose and hexose produced by the two enzymes were purified and structurally identified. Molecular docking with κ-carrageenan hexose suggested that the efficiency improvement after truncation might be attributed to the conformation differences of the two enzymes.

Novel Salt-Tolerant Xylanase from a Mangrove-Isolated Fungus Phoma sp. MF13 and Its Application in Chinese Steamed Bread.

A novel glycosyl hydrolase family 11 xylanase gene, xynMF13A, was cloned from Phoma sp. MF13, a xylanase-producing fungus isolated from mangrove sediment. xynMF13A was heterologously expressed in Pichia pastoris, and the recombinant XynMF13A (rXynMF13A) was purified by Ni-affinity chromatography. The temperature and pH optima of purified rXynMF13A were 45 °C and pH 5.0, respectively. rXynMF13A showed a high level of salt tolerance, with maximal enzyme activity being seen at 0.5 M NaCl and as much as 53% of maximal activity at 4 M NaCl. The major rXynMF13A hydrolysis products from corncob xylan were xylobiose, xylotriose, xylotetraose, and xylopentaose, but no xylose was found. These hydrolysis products suggest an important potential for XynMF13A in the production of xylooligosaccharides (XOs). Furthermore, rXynMF13A had beneficial effects on Chinese steamed bread production, by increasing specific volume and elasticity while decreasing hardness and chewiness. These results demonstrate XynMF13A to be a novel xylanase with potentially significant applications in baking, XOs production, and seafood processing.

RFAthM6A: a new tool for predicting m6A sites in Arabidopsis thaliana.

KEY MESSAGE: We curated a reliable dataset of m6A sites in Arabidopsis thaliana, built competitive models for predicting m6A sites, extracted predominant rules from the prediction models and analyzed the most important features. In biological RNA, approximately 150 chemical modifications have been discovered, of which N6-methyladenine (m6A) is the most prevalent and abundant. This modification plays an essential role in a myriad of biological mechanisms and regulates RNA localization, nuclear export, translation, stability, alternative splicing, and other processes. However, m6A-seq and other wet-lab techniques do not easily facilitate accurate and complete determination of m6A sites across the transcriptome. Therefore, the use of computational methods to establish accurate models for predicting m6A sites is essential. In this work, we manually curated a reliable dataset of m6A sites and non-m6A sites and developed a new tool called RFAthM6A for predicting m6A sites in Arabidopsis thaliana. Briefly, RFAthM6A consists of four independent models named RFPSNSP, RFPSDSP, RFKSNPF and RFKNF and strict benchmarks show that the AUC values of the four models reached 0.894, 0.914, 0.920 and 0.926, respectively in a fivefold cross validation and the prediction performance of RFPSDSP, RFKSNPF and RFKNF exceeded that of three previously reported models (AthMethPre, M6ATH and RAM-NPPS). Linear combination of the prediction scores of RFPSDSP, RFKSNPF and RFKNF improved the prediction performance. We also extracted several predominant rules that underlie the m6A site identification from the trained models. Furthermore, the most important features of the predictors for the m6A site identification were also analyzed in depth. To facilitate use of our proposed models by interested researchers, all the source codes and datasets are publicly deposited at https://github.com/nongdaxiaofeng/RFAthM6A .

SOHPRED: a new bioinformatics tool for the characterization and prediction of human S-sulfenylation sites.

Protein S-sulfenylation (SOH) is a type of post-translational modification through the oxidation of cysteine thiols to sulfenic acids. It acts as a redox switch to modulate versatile cellular processes and plays important roles in signal transduction, protein folding and enzymatic catalysis. Reversible SOH is also a key component for maintaining redox homeostasis and has been implicated in a variety of human diseases, such as cancer, diabetes, and atherosclerosis, due to redox imbalance. Despite its significance, the in situ trapping of the entire 'sulfenome' remains a major challenge. Yang et al. have recently experimentally identified about 1000 SOH sites, providing an enriched benchmark SOH dataset. In this work, we developed a new ensemble learning tool SOHPRED for identifying protein SOH sites based on the compositions of enriched amino acids and the physicochemical properties of residues surrounding SOH sites. SOHPRED was built based on four complementary predictors, i.e. a naive Bayesian predictor, a random forest predictor and two support vector machine predictors, whose training features are, respectively, amino acid occurrences, physicochemical properties, frequencies of k-spaced amino acid pairs and sequence profiles. Benchmarking experiments on the 5-fold cross validation and independent tests show that SOHPRED achieved AUC values of 0.784 and 0.799, respectively, which outperforms several previously developed tools. As a real application of SOHPRED, we predicted potential SOH sites for 193 S-sulfenylated substrates, which had been experimentally detected through a global sulfenome profiling in living cells, though the actual SOH sites were not determined. The web server of SOHPRED has been made publicly available at for the wider research community. The source codes and the benchmark datasets can be downloaded from the website.

DephosSite: a machine learning approach for discovering phosphotase-specific dephosphorylation sites.

Protein dephosphorylation, which is an inverse process of phosphorylation, plays a crucial role in a myriad of cellular processes, including mitotic cycle, proliferation, differentiation, and cell growth. Compared with tyrosine kinase substrate and phosphorylation site prediction, there is a paucity of studies focusing on computational methods of predicting protein tyrosine phosphatase substrates and dephosphorylation sites. In this work, we developed two elegant models for predicting the substrate dephosphorylation sites of three specific phosphatases, namely, PTP1B, SHP-1, and SHP-2. The first predictor is called MGPS-DEPHOS, which is modified from the GPS (Group-based Prediction System) algorithm with an interpretable capability. The second predictor is called CKSAAP-DEPHOS, which is built through the combination of support vector machine (SVM) and the composition of k-spaced amino acid pairs (CKSAAP) encoding scheme. Benchmarking experiments using jackknife cross validation and 30 repeats of 5-fold cross validation tests show that MGPS-DEPHOS and CKSAAP-DEPHOS achieved AUC values of 0.921, 0.914 and 0.912, for predicting dephosphorylation sites of the three phosphatases PTP1B, SHP-1, and SHP-2, respectively. Both methods outperformed the previously developed kNN-DEPHOS algorithm. In addition, a web server implementing our algorithms is publicly available at http://genomics.fzu.edu.cn/dephossite/ for the research community.

A novel cold-adapted and highly salt-tolerant esterase from Alkalibacterium sp. SL3 from the sediment of a soda lake.

A novel esterase gene (estSL3) was cloned from the Alkalibacterium sp. SL3, which was isolated from the sediment of soda lake Dabusu. The 636-bp full-length gene encodes a polypeptide of 211 amino acid residues that is closely related with putative GDSL family lipases from Alkalibacterium and Enterococcus. The gene was successfully expressed in E. coli, and the recombinant protein (rEstSL3) was purified to electrophoretic homogeneity and characterized. rEstSL3 exhibited the highest activity towards pNP-acetate and had no activity towards pNP-esters with acyl chains longer than C8. The enzyme was highly cold-adapted, showing an apparent temperature optimum of 30 °C and remaining approximately 70% of the activity at 0 °C. It was active and stable over the pH range from 7 to 10, and highly salt-tolerant up to 5 M NaCl. Moreover, rEstSL3 was strongly resistant to most tested metal ions, chemical reagents, detergents and organic solvents. Amino acid composition analysis indicated that EstSL3 had fewer proline residues, hydrogen bonds and salt bridges than mesophilic and thermophilic counterparts, but more acidic amino acids and less hydrophobic amino acids when compared with other salt-tolerant esterases. The cold active, salt-tolerant and chemical-resistant properties make it a promising enzyme for basic research and industrial applications.

A Novel Multi-domain High Molecular, Salt-Stable Alkaline Xylanase from Alkalibacterium sp. SL3.

A novel multi-domain high molecular xylanase coding gene (xynSL3) was cloned from Alkalibacterium sp. SL3, an alkaliphilic bacterial strain isolated from the sediment of soda lake Dabusu. The deduced XynSL3 is composed of a putative signal peptide, three tandem domains of carbohydrate binding module (CBM) family 22, a catalytic domain of glycosyl hydrolase (GH) family 10 and a domain of CBM9. XynSL3 shares the highest identity of 66% to a hypothetical protein from Alkalibacterium sp. AK22 and has low identities (33-45%) with other functionally characterized xylanases. The gene xynSL3 was expressed heterologously in Escherichia coli and the recombinant enzyme demonstrated some particular characteristics. Purified recombinant XynSL3 (rXynSL3) was highly active and stable over the neutral and alkaline pH ranges from 7.0 to 12.0, with maximum activity at pH 9.0 and around 45% activity at pH 12.0. It had an apparent temperature optimum of 55°C and was stable at 50°C. The rXynSL3 was highly halotolerant, retaining more than 60% activity with 3 M NaCl and was stable at up to a 4 M concentration of NaCl. The hydrolysis products of rXynSL3 from corncob xylan were mainly xylobiose and xylotetraose. The activity of rXynSL3 was enhanced by Ca2+ and it has strong resistance to sodium dodecyl sulfate (SDS). This multi-domain, alkaline and salt-tolerant enzyme has great potential for basic research and industrial applications such as the biobleaching of paper pulp and production of xylo-oligosaccharides (XOS).

Prediction of structural features and application to outer membrane protein identification.

Protein three-dimensional (3D) structures provide insightful information in many fields of biology. One-dimensional properties derived from 3D structures such as secondary structure, residue solvent accessibility, residue depth and backbone torsion angles are helpful to protein function prediction, fold recognition and ab initio folding. Here, we predict various structural features with the assistance of neural network learning. Based on an independent test dataset, protein secondary structure prediction generates an overall Q3 accuracy of ~80%. Meanwhile, the prediction of relative solvent accessibility obtains the highest mean absolute error of 0.164, and prediction of residue depth achieves the lowest mean absolute error of 0.062. We further improve the outer membrane protein identification by including the predicted structural features in a scoring function using a simple profile-to-profile alignment. The results demonstrate that the accuracy of outer membrane protein identification can be improved by ~3% at a 1% false positive level when structural features are incorporated. Finally, our methods are available as two convenient and easy-to-use programs. One is PSSM-2-Features for predicting secondary structure, relative solvent accessibility, residue depth and backbone torsion angles, the other is PPA-OMP for identifying outer membrane proteins from proteomes.

AAFreqCoil: a new classifier to distinguish parallel dimeric and trimeric coiled coils.

Coiled coils are characteristic rope-like protein structures, constituted by one or more heptad repeats. Native coiled-coil structures play important roles in various biological processes, while the designed ones are widely employed in medicine and industry. To date, two major oligomeric states (i.e. dimeric and trimeric states) of a coiled-coil structure have been observed, plausibly exerting different biological functions. Therefore, exploration of the relationship between heptad repeat sequences and coiled coil structures is highly important. In this paper, we develop a new method named AAFreqCoil to classify parallel dimeric and trimeric coiled coils. Our method demonstrated its competitive performance when benchmarked based on 10-fold cross validation and jackknife cross validation. Meanwhile, the rules that can explicitly explain the prediction results of the test coiled coil can be extracted from the AAFreqCoil model for a better explanation of user predictions. A web server and stand-alone program implementing the AAFreqCoil algorithm are freely available at .

GPCRserver: an accurate and novel G protein-coupled receptor predictor.

G protein coupled receptors (GPCRs), also known as seven-transmembrane domain receptors, pass through the cellular membrane seven times and play diverse biological roles in the cells such as signaling, transporting of molecules and cell-cell communication. In this work, we develop a web server, namely the GPCRserver, which is capable of identifying GPCRs from genomic sequences, and locating their transmembrane regions. The GPCRserver contains three modules: (1) the Trans-GPCR for the transmembrane region prediction by using sequence evolutionary profiles with the assistance of neural network training, (2) the SSEA-GPCR for identifying GPCRs from genomic data by using secondary structure element alignment, and (3) the PPA-GPCR for identifying GPCRs by using profile-to-profile alignment. Our predictor was strictly benchmarked and showed its favorable performance in the real application. The web server and stand-alone programs are publicly available at .