T4SEfinder: a bioinformatics tool for genome-scale prediction of bacterial type IV secreted effectors using pre-trained protein language model

T4SEpp: A pipeline integrating protein language models to predict bacterial type IV secreted effectors

Many pathogenic bacteria use type IV secretion systems (T4SSs) to deliver effectors (T4SEs) into the cytoplasm of eukaryotic cells, causing diseases. The identification of effectors is a crucial step in understanding the mechanisms of bacterial pathogenicity, but this remains a major challenge. In this study, we used the full-length embedding features generated by six pre-trained protein language models to train classifiers predicting T4SEs and compared their performance. We integrated three modules into a model called T4SEpp. The first module searched for full-length homologs of known T4SEs, signal sequences, and effector domains; the second module fine-tuned a machine learning model using data for a signal sequence feature; and the third module used the three best-performing pre-trained protein language models. T4SEpp outperformed other state-of-the-art (SOTA) software tools, achieving ∼0.98 accuracy at a high specificity of ∼0.99, based on the assessment of an independent validation dataset. T4SEpp predicted 13 T4SEs from Helicobacter pylori, including the well-known CagA and 12 other potential ones, among which eleven could potentially interact with human proteins. This suggests that these potential T4SEs may be associated with the pathogenicity of H. pylori. Overall, T4SEpp provides a better solution to assist in the identification of bacterial T4SEs and facilitates studies of bacterial pathogenicity. T4SEpp is freely accessible at https://bis.zju.edu.cn/T4SEpp.

The F plasmid conjutome: the repertoire of E. coli proteins translocated through an F-encoded type IV secretion system

Bacterial conjugation systems pose a major threat to human health through their widespread dissemination of mobile genetic elements (MGEs) carrying cargoes of antibiotic resistance genes. Using the Cre Recombinase Assay for Translocation (CRAfT), we recently reported that the IncFV pED208 conjugation system also translocates at least 16 plasmid-encoded proteins to recipient bacteria. Here, we deployed a high-throughput CRAfT screen to identify the repertoire of chromosomally encoded protein substrates of the pED208 system. We identified 32 substrates encoded by the Escherichia coli W3110 genome with functions associated with (i) DNA/nucleotide metabolism, (ii) stress tolerance/physiology, (iii) transcriptional regulation, or (iv) toxin inhibition. The respective gene deletions did not impact pED208 transfer proficiencies, nor did Group 1 (DNA/nucleotide metabolism) mutations detectably alter the SOS response elicited in new transconjugants upon acquisition of pED208. However, MC4100(pED208) donor cells intrinsically exhibit significantly higher SOS activation than plasmid-free MC4100 cells, and this plasmid carriage-induced stress response is further elevated in donor cells deleted of several Group 1 genes. Among 10 characterized substrates, we gained evidence of C-terminal or internal translocation signals that could function independently or synergistically for optimal protein transfer. Remarkably, nearly all tested proteins were also translocated through the IncN pKM101 and IncP RP4 conjugation systems. This repertoire of E. coli protein substrates, here termed the F plasmid "conjutome," is thus characterized by functions of potential benefit to new transconjugants, diverse TSs, and the capacity for promiscuous transfer through heterologous conjugation systems.

IMPORTANCE

Conjugation systems comprise a major subfamily of the type IV secretion systems (T4SSs) and are the progenitors of a second large T4SS subfamily dedicated to translocation of protein effectors. This study examined the capacity of conjugation machines to function as protein translocators. Using a high-throughput reporter screen, we determined that 32 chromosomally encoded proteins are delivered through an F plasmid conjugation system. The translocated proteins potentially enhance the establishment of the co-transferred F plasmid or mitigate mating-induced stresses. Translocation signals located C-terminally or internally conferred substrate recognition by the F system and, remarkably, many substrates also were translocated through heterologous conjugation systems. Our findings highlight the plasticity of conjugation systems in their capacities to co-translocate DNA and many protein substrates.

CELA-MFP: a contrast-enhanced and label-adaptive framework for multi-functional therapeutic peptides prediction

Functional peptides play crucial roles in various biological processes and hold significant potential in many fields such as drug discovery and biotechnology. Accurately predicting the functions of peptides is essential for understanding their diverse effects and designing peptide-based therapeutics. Here, we propose CELA-MFP, a deep learning framework that incorporates feature Contrastive Enhancement and Label Adaptation for predicting Multi-Functional therapeutic Peptides. CELA-MFP utilizes a protein language model (pLM) to extract features from peptide sequences, which are then fed into a Transformer decoder for function prediction, effectively modeling correlations between different functions. To enhance the representation of each peptide sequence, contrastive learning is employed during training. Experimental results demonstrate that CELA-MFP outperforms state-of-the-art methods on most evaluation metrics for two widely used datasets, MFBP and MFTP. The interpretability of CELA-MFP is demonstrated by visualizing attention patterns in pLM and Transformer decoder. Finally, a user-friendly online server for predicting multi-functional peptides is established as the implementation of the proposed CELA-MFP and can be freely accessed at http://dreamai.cmii.online/CELA-MFP.

Single-cell genomics revealed Candidatus Grellia alia sp. nov. as an endosymbiont of Eutreptiella sp. (Euglenophyceae)

Though endosymbioses between protists and prokaryotes are widespread, certain host lineages have received disproportionate attention what may indicate either a predisposition to such interactions or limited studies on certain protist groups due to lack of cultures. The euglenids represent one such group in spite of microscopic observations showing intracellular bacteria in some strains. Here, we perform a comprehensive molecular analysis of a previously identified endosymbiont in the Eutreptiella sp. CCMP3347 using a single cell approach and bulk culture sequencing. The genome reconstruction of this endosymbiont allowed the description of a new endosymbiont Candidatus Grellia alia sp. nov. from the family Midichloriaceae. Comparative genomics revealed a remarkably complete conjugative type IV secretion system present in three copies on the plasmid sequences of the studied endosymbiont, a feature missing in the closely related Grellia incantans. This study addresses the challenge of limited host cultures with endosymbionts by showing that the genomes of endosymbionts reconstructed from single host cells have the completeness and contiguity that matches or exceeds those coming from bulk cultures. This paves the way for further studies of endosymbionts in euglenids and other protist groups. The research also provides the opportunity to study the diversity of endosymbionts in natural populations.

POOE: predicting oomycete effectors based on a pre-trained large protein language model

Oomycetes are fungus-like eukaryotic microorganisms which can cause catastrophic diseases in many plants. Successful infection of oomycetes depends highly on their effector proteins that are secreted into plant cells to subvert plant immunity. Thus, systematic identification of effectors from the oomycete proteomes remains an initial but crucial step in understanding plant-pathogen relationships. However, the number of experimentally identified oomycete effectors is still limited. Currently, only a few bioinformatics predictors exist to detect potential effectors, and their prediction performance needs to be improved. Here, we used the sequence embeddings from a pre-trained large protein language model (ProtTrans) as input and developed a support vector machine-based method called POOE for predicting oomycete effectors. POOE could achieve a highly accurate performance with an area under the precision-recall curve of 0.804 (area under the receiver operating characteristic curve = 0.893, accuracy = 0.874, precision = 0.777, recall = 0.684, and specificity = 0.936) in the fivefold cross-validation, considerably outperforming various combinations of popular machine learning algorithms and other commonly used sequence encoding schemes. A similar prediction performance was also observed in the independent test. Compared with the existing oomycete effector prediction methods, POOE provided very competitive and promising performance, suggesting that ProtTrans effectively captures rich protein semantic information and dramatically improves the prediction task. We anticipate that POOE can accelerate the identification of oomycete effectors and provide new hints to systematically understand the functional roles of effectors in plant-pathogen interactions. The web server of POOE is freely accessible at http://zzdlab.com/pooe/index.php. The corresponding source codes and data sets are also available at https://github.com/zzdlabzm/POOE.IMPORTANCEIn this work, we use the sequence representations from a pre-trained large protein language model (ProtTrans) as input and develop a Support Vector Machine-based method called POOE for predicting oomycete effectors. POOE could achieve a highly accurate performance in the independent test set, considerably outperforming existing oomycete effector prediction methods. We expect that this new bioinformatics tool will accelerate the identification of oomycete effectors and further guide the experimental efforts to interrogate the functional roles of effectors in plant-pathogen interaction.

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.

DeepSecE: A Deep-Learning-Based Framework for Multiclass Prediction of Secreted Proteins in Gram-Negative Bacteria

Proteins secreted by Gram-negative bacteria are tightly linked to the virulence and adaptability of these microbes to environmental changes. Accurate identification of such secreted proteins can facilitate the investigations of infections and diseases caused by these bacterial pathogens. However, current bioinformatic methods for predicting bacterial secreted substrate proteins have limited computational efficiency and application scope on a genome-wide scale. Here, we propose a novel deep-learning-based framework-DeepSecE-for the simultaneous inference of multiple distinct groups of secreted proteins produced by Gram-negative bacteria. DeepSecE remarkably improves their classification from nonsecreted proteins using a pretrained protein language model and transformer, achieving a macro-average accuracy of 0.883 on 5-fold cross-validation. Performance benchmarking suggests that DeepSecE achieves competitive performance with the state-of-the-art binary predictors specialized for individual types of secreted substrates. The attention mechanism corroborates salient patterns and motifs at the N or C termini of the protein sequences. Using this pipeline, we further investigate the genome-wide prediction of novel secreted proteins and their taxonomic distribution across ~1,000 Gram-negative bacterial genomes. The present analysis demonstrates that DeepSecE has major potential for the discovery of disease-associated secreted proteins in a diverse range of Gram-negative bacteria. An online web server of DeepSecE is also publicly available to predict and explore various secreted substrate proteins via the input of bacterial genome sequences.

Identification and Preliminary Characterization of Novel Type III Secreted Effector Proteins in Chlamydia trachomatis

Chlamydia trachomatis is an obligate intracellular pathogen that replicates in a host-derived vacuole termed the inclusion. Central to pathogenesis is a type III secretion system that translocates effector proteins into the host cell, which are predicted to play major roles in host cell invasion, nutrient acquisition, and immune evasion. However, until recently, the genetic intractability of C. trachomatis hindered identification and characterization of these important virulence factors. Here, we sought to expand the repertoire of identified effector proteins and confirm they are secreted during C. trachomatis infection. Utilizing bioinformatics, we identified 18 candidate substrates that had not been previously assessed for secretion, of which we show four to be secreted, using Yersinia pseudotuberculosis as a surrogate host. Using adenylate cyclase (CyaA), BlaM, and glycogen synthase kinase (GSK) secretion assays, we identified nine novel substrates that were secreted in at least one assay. Interestingly, only three of the substrates, shown to be translocated by C. trachomatis, were similarly secreted by Y. pseudotuberculosis. Using large-scale screens to determine subcellular localization and identify effectors that perturb crucial host cell processes, we identified one novel substrate, CT392, that is toxic when heterologously expressed in Saccharomyces cerevisiae. Toxicity required both the N- and C-terminal regions of the protein. Additionally, we show that these newly described substrates traffic to distinct host cell compartments, including vesicles and the cytoplasm. Collectively, our study expands the known repertoire of C. trachomatis secreted factors and highlights the importance of testing for secretion in the native host using multiple secretion assays when possible.

Comparative genomics and virulome analysis reveal unique features associated with clinical strains of Klebsiella pneumoniae and Klebsiella quasipneumoniae from Trinidad, West Indies

Klebsiella pneumoniae and Klebsiella quasipneumoniae are closely related human pathogens of global concern. The more recently described K. quasipneumoniae shares similar morphological characteristics with K. pneumoniae and is commonly misidentified as this species using traditional laboratory techniques. The vast mobilome in these pathogenic bacteria influences the dissemination of virulence factors in high-risk environments and it is, therefore, critical to monitor strains for developing effective clinical management strategies. Herein, this study utilized Illumina sequencing to characterize the whole genomes of nine clinical K. pneumoniae and one K. quasipneumoniae isolate obtained from patients of 3 major hospitals in Trinidad, West Indies. Reconstruction of the assembled genomes and implementation of several bioinformatic tools revealed unique features such as high pathogenicity islands associated with the isolates. The K. pneumoniae isolates were categorized as classical (n = 3), uropathogenic (n = 5), or hypervirulent (n = 1) strains. In silico multilocus sequence typing, and phylogenetic analysis showed that isolates were related to several international high-risk genotypes, including sequence types ST11, ST15, ST86, and ST307. Analysis of the virulome and mobilome of these pathogens showed unique and clinically important features including the presence of genes associated with Type 1 and Type 3 fimbriae, the aerobactin and yersiniabactin siderophore systems, the K2 and O1/2, and the O3 and O5 serotypes. These genes were either on or in close proximity to insertion sequence elements, phage sequences, and plasmids. Several secretion systems including the Type VI system and relevant effector proteins were prevalent in the local isolates. This is the first comprehensive study investigating the genomes of clinical K. pneumoniae and K. quasipneumoniae isolates from Trinidad, West Indies. The data presented illustrate the diversity of Trinidadian clinical K. pneumoniae isolates as well as significant virulence biomarkers and mobile elements associated with these isolates. Additionally, the genomes of the local isolates will add to global databases and thus can be used in future surveillance or genomic studies in this country and the wider Caribbean region.

Genomic Assessment of the Contribution of the Wolbachia Endosymbiont of Eurosta solidaginis to Gall Induction

We explored the genome of the Wolbachia strain, wEsol, symbiotic with the plant-gall-inducing fly Eurosta solidaginis with the goal of determining if wEsol contributes to gall induction by its insect host. Gall induction by insects has been hypothesized to involve the secretion of the phytohormones cytokinin and auxin and/or proteinaceous effectors to stimulate cell division and growth in the host plant. We sequenced the metagenome of E. solidaginis and wEsol and assembled and annotated the genome of wEsol. The wEsol genome has an assembled length of 1.66 Mbp and contains 1878 protein-coding genes. The wEsol genome is replete with proteins encoded by mobile genetic elements and shows evidence of seven different prophages. We also detected evidence of multiple small insertions of wEsol genes into the genome of the host insect. Our characterization of the genome of wEsol indicates that it is compromised in the synthesis of dimethylallyl pyrophosphate (DMAPP) and S-adenosyl L-methionine (SAM), which are precursors required for the synthesis of cytokinins and methylthiolated cytokinins. wEsol is also incapable of synthesizing tryptophan, and its genome contains no enzymes in any of the known pathways for the synthesis of indole-3-acetic acid (IAA) from tryptophan. wEsol must steal DMAPP and L-methionine from its host and therefore is unlikely to provide cytokinin and auxin to its insect host for use in gall induction. Furthermore, in spite of its large repertoire of predicted Type IV secreted effector proteins, these effectors are more likely to contribute to the acquisition of nutrients and the manipulation of the host's cellular environment to contribute to growth and reproduction of wEsol than to aid E. solidaginis in manipulating its host plant. Combined with earlier work that shows that wEsol is absent from the salivary glands of E. solidaginis, our results suggest that wEsol does not contribute to gall induction by its host.

Identification and characterization of opportunistic pathogen Pectobacterium polonicum causing potato blackleg in China

Blackleg and aerial stem rot of potato (Solanum tuberosum L.), caused by soft rot enterobacteria of the genera Pectobacterium and Dickeya, has recently increased years in Hebei Province, China. Field surveys were performed during the 2021 potato growing season in Hebei to identify and characterize bacterial pathogens. Sixteen potato plants showing blackleg or aerial stem rot were collected from three potato-producing areas, and ten representative pectinolytic bacteria were isolated from symptomatic plants. 16S rDNA sequencing and multilocus sequence analysis were performed to determine the taxonomic position of the bacterial isolates. The isolates belonged to the genus Pectobacterium, including Pectobacterium atrosepticum, Pectobacterium carotovorum, Pectobacterium brasiliense, and Pectobacterium parmentieri. The exceptions were isolates BY21311 and BY21312, which belonged to a new species of Pectobacterium polonicum previously found in groundwater. The taxonomy of isolate BY21311 was confirmed using whole genome-based analysis. P. polonicum has only been identified in potato plants on one farm in Baoding region in China. Isolates BY21311 and BY21312 displayed similar physiological and biochemical traits to the type strain DPMP315^T. Artificial inoculation assays revealed that isolate BY21311 fulfilled Koch's postulates for potato blackleg. These findings represent the first time P. polonicum, a water-associated Pectobacterium species may be the cause of blackleg in the field. Interestingly, P. polonicum BY21311 has reduced ability to macerate potato tubers when compared to P. atrosepticum, P. brasiliense, P. versatile, and P. parvum, which is more virulent in tubers than the type strain DPMP315^T. The host range of isolate BY21311 was determined by injection method, which can impregnate five plants. Although the genome of isolate BY21311 harbors gene clusters encoding a type III secretion system, it did not elicit a hypersensitive response (HR) in Nicotiana benthamiana or N. tabacum leaves. T3SS effector AvrE and T4SS effector PilN were obtained by predicting isolate BY21311 genome. P. polonicum appears to show significant variations in gene content between two genomes, and gene content varies between isolates BY21311 and DPMP315^T, with strain specific-genes involved in many aspects, including lipopolysaccharide biosynthesis, substrate translocation, T4SS and T6SS among others, suggesting that isolates BY21311 and DPMP315^T might represent distinct clades within the species.

AcneTyper: An automatic diagnosis method of dermoscopic acne image via self-ensemble and stacking

BACKGROUND

Acne is a skin lesion type widely existing in adolescents, and poses computational challenges for automatic diagnosis. Computer vision algorithms are utilized to detect and determine different subtypes of acne. Most of the existing acne detection algorithms are based on the facial natural images, which carry noisy factors like illuminations.

OBJECTIVE

In order to tackle this issue, this study collected a dataset ACNEDer of dermoscopic acne images with annotations. Deep learning methods have demonstrated powerful capabilities in automatic acne diagnosis, and they usually release the training epoch with the best performance as the delivered model.

METHODS

This study proposes a novel self-ensemble and stacking-based framework AcneTyper for diagnosing the acne subtypes. Instead of delivering the best epoch, AcneTyper consolidates the prediction results of all training epochs as the latent features and stacks the best subset of these latent features for distinguishing different acne subtypes.

RESULTS

The proposed AcneTyper framework achieves a promising detection performance of acne subtypes and even outperforms a clinical dermatologist with two-year experiences by 6.8% in accuracy.

CONCLUSION

The method we proposed is used to determine different subtypes of acne and outperforms inexperienced dermatologists and contributes to reducing the probability of misdiagnosis.