Balrog: A universal protein model for prokaryotic gene prediction.
Ontology highlight
ABSTRACT:
SUBMITTER: Sommer MJ
PROVIDER: S-EPMC7946324 | biostudies-literature | 2021 Feb
REPOSITORIES: biostudies-literature
Publications
Balrog: A universal protein model for prokaryotic gene prediction.
Sommer Markus J MJ Salzberg Steven L SL
PLoS computational biology 20210226 2
Low-cost, high-throughput sequencing has led to an enormous increase in the number of sequenced microbial genomes, with well over 100,000 genomes in public archives today. Automatic genome annotation tools are integral to understanding these organisms, yet older gene finding methods must be retrained on each new genome. We have developed a universal model of prokaryotic genes by fitting a temporal convolutional network to amino-acid sequences from a large, diverse set of microbial genomes. We in ...[more]
Similar Datasets
Project description:With almost no consensus promoter sequence in prokaryotes, recruitment of RNA polymerase (RNAP) to precise transcriptional start sites (TSSs) has remained an unsolved puzzle. Uncovering the underlying mechanism is critical for understanding the principle of gene regulation. We attempted to search the hidden code in ∼16,500 promoters of 12 prokaryotes representing two kingdoms in their structure and energetics. Twenty-eight fundamental parameters of DNA structure including backbone angles, basepair axis, and interbasepair and intrabasepair parameters were used, and information was extracted from x-ray crystallography data. Three parameters (solvation energy, hydrogen-bond energy, and stacking energy) were selected for creating energetics profiles using in-house programs. DNA of promoter regions was found to be inherently designed to undergo a change in every parameter undertaken for the study, in all prokaryotes. The change starts from some distance upstream of TSSs and continues past some distance from TSS, hence giving a signature state to promoter regions. These signature states might be the universal hidden codes recognized by RNAP. This observation was reiterated when randomly selected promoter sequences (with little sequence conservation) were subjected to structure generation; all developed into very similar three-dimensional structures quite distinct from those of conventional B-DNA and coding sequences. Fine structural details at important motifs (viz. -11, -35, and -75 positions relative to TSS) of promoters reveal novel to our knowledge and pointed insights for RNAP interaction at these locations; it could be correlated with how some particular structural changes at the -11 region may allow insertion of RNAP amino acids in interbasepair space as well as facilitate the flipping out of bases from the DNA duplex.
| S-EPMC6170797 | biostudies-literature
Project description:BackgroundViruses are ubiquitous biological entities, estimated to be the largest reservoirs of unexplored genetic diversity on Earth. Full functional characterization and annotation of newly discovered viruses requires tools to enable taxonomic assignment, the range of hosts, and biological properties of the virus. Here we focus on prokaryotic viruses, which include phages and archaeal viruses, and for which identifying the viral host is an essential step in characterizing the virus, as the virus relies on the host for survival. Currently, the method for determining the viral host is either to culture the virus, which is low-throughput, time-consuming, and expensive, or to computationally predict the viral hosts, which needs improvements at both accuracy and usability. Here we develop a Gaussian model to predict hosts for prokaryotic viruses with better performances than previous computational methods.ResultsWe present here Prokaryotic virus Host Predictor (PHP), a software tool using a Gaussian model, to predict hosts for prokaryotic viruses using the differences of k-mer frequencies between viral and host genomic sequences as features. PHP gave a host prediction accuracy of 34% (genus level) on the VirHostMatcher benchmark dataset and a host prediction accuracy of 35% (genus level) on a new dataset containing 671 viruses and 60,105 prokaryotic genomes. The prediction accuracy exceeded that of two alignment-free methods (VirHostMatcher and WIsH, 28-34%, genus level). PHP also outperformed these two alignment-free methods much (24-38% vs 18-20%, genus level) when predicting hosts for prokaryotic viruses which cannot be predicted by the BLAST-based or the CRISPR-spacer-based methods alone. Requiring a minimal score for making predictions (thresholding) and taking the consensus of the top 30 predictions further improved the host prediction accuracy of PHP.ConclusionsThe Prokaryotic virus Host Predictor software tool provides an intuitive and user-friendly API for the Gaussian model described herein. This work will facilitate the rapid identification of hosts for newly identified prokaryotic viruses in metagenomic studies.
| S-EPMC7807511 | biostudies-literature
Project description:State-of-the-art algorithms of ab initio gene prediction for prokaryotic genomes were shown to be sufficiently accurate. A pair of algorithms would agree on predictions of gene 3'ends. Nonetheless, predictions of gene starts would not match for 15-25% of genes in a genome. This discrepancy is a serious issue that is difficult to be resolved due to the absence of sufficiently large sets of genes with experimentally verified starts. We have introduced StartLink that infers gene starts from conservation patterns revealed by multiple alignments of homologous nucleotide sequences. We also have introduced StartLink+ combining both ab initio and alignment-based methods. The ability of StartLink to predict the start of a given gene is restricted by the availability of homologs in a database. We observed that StartLink made predictions for 85% of genes per genome on average. The StartLink+ accuracy was shown to be 98-99% on the sets of genes with experimentally verified starts. In comparison with database annotations, we observed that the annotated gene starts deviated from the StartLink+ predictions for ∼5% of genes in AT-rich genomes and for 10-15% of genes in GC-rich genomes on average. The use of StartLink+ has a potential to significantly improve gene start annotation in genomic databases.
| S-EPMC9581028 | biostudies-literature
Project description:In the prokaryotic kingdom, protein phosphorylation serves as one of the most important posttranslational modifications (PTMs) and is involved in orchestrating a broad spectrum of biological processes. Here, we report an updated online server named the group-based prediction system for prokaryotic phosphorylation language model (GPS-pPLM), used for predicting phosphorylation sites (p-sites) in prokaryotes. For model training, two deep learning methods, a transformer and a deep neural network, were employed, and a total of 10 sequence features and contextual features were integrated. Using 44,839 nonredundant p-sites in 16,041 proteins from 95 prokaryotes, two general models for the prediction of O-phosphorylation and N-phosphorylation were first pretrained and then fine-tuned to construct 6 predictors specific for each phosphorylatable residue type as well as 134 species-specific predictors. Compared with other existing tools, the GPS-pPLM exhibits higher accuracy in predicting prokaryotic O-phosphorylation p-sites. Protein sequences in FASTA format or UniProt accession numbers can be submitted by users, and the predicted results are displayed in tabular form. In addition, we annotate the predicted p-sites with knowledge from 22 public resources, including experimental evidence, 3D structures, and disorder tendencies. The online service of the GPS-pPLM is freely accessible for academic research.
| S-EPMC11593113 | biostudies-literature
Project description:Identification of prokaryotic transposases (Tnps) not only gives insight into the spread of antibiotic resistance and virulence but the process of DNA movement. This study aimed to develop a classifier for predicting Tnps in bacteria and archaea using machine learning (ML) approaches. We extracted a total of 2751 protein features from the training dataset including 14852 Tnps and 14852 controls, and selected 75 features as predictive signatures using the combined mutual information and least absolute shrinkage and selection operator algorithms. By aggregating these signatures, an ensemble classifier that integrated a collection of individual ML-based classifiers, was developed to identify Tnps. Further validation revealed that this classifier achieved good performance with an average AUC of 0.955, and met or exceeded other common methods. Based on this ensemble classifier, a stand-alone command-line tool designated TnpDiscovery was established to maximize the convenience for bioinformaticians and experimental researchers toward Tnp prediction. This study demonstrates the effectiveness of ML approaches in identifying Tnps, facilitating the discovery of novel Tnps in the future.
| S-EPMC8477400 | biostudies-literature
Project description:BackgroundImmunotherapy is an emerging approach in cancer treatment that activates the host immune system to destroy cancer cells expressing unique peptide signatures (neoepitopes). Administrations of cancer-specific neoepitopes in the form of synthetic peptide vaccine have been proven effective in both mouse models and human patients. Because only a tiny fraction of cancer-specific neoepitopes actually elicits immune response, selection of potent, immunogenic neoepitopes remains a challenging step in cancer vaccine development. A basic approach for immunogenicity prediction is based on the premise that effective neoepitope should bind with the Major Histocompatibility Complex (MHC) with high affinity.ResultsIn this study, we developed MHCSeqNet, an open-source deep learning model, which not only outperforms state-of-the-art predictors on both MHC binding affinity and MHC ligand peptidome datasets but also exhibits promising generalization to unseen MHC class I alleles. MHCSeqNet employed neural network architectures developed for natural language processing to model amino acid sequence representations of MHC allele and epitope peptide as sentences with amino acids as individual words. This consideration allows MHCSeqNet to accept new MHC alleles as well as peptides of any length.ConclusionsThe improved performance and the flexibility offered by MHCSeqNet should make it a valuable tool for screening effective neoepitopes in cancer vaccine development.
| S-EPMC6540523 | biostudies-literature
Project description:Computationally annotating proteins with a molecular function is a difficult problem that is made even harder due to the limited amount of available labeled protein training data. Unsupervised protein embeddings partly circumvent this limitation by learning a universal protein representation from many unlabeled sequences. Such embeddings incorporate contextual information of amino acids, thereby modeling the underlying principles of protein sequences insensitive to the context of species. We used an existing pre-trained protein embedding method and subjected its molecular function prediction performance to detailed characterization, first to advance the understanding of protein language models, and second to determine areas of improvement. Then, we applied the model in a transfer learning task by training a function predictor based on the embeddings of annotated protein sequences of one training species and making predictions on the proteins of several test species with varying evolutionary distance. We show that this approach successfully generalizes knowledge about protein function from one eukaryotic species to various other species, outperforming both an alignment-based and a supervised-learning-based baseline. This implies that such a method could be effective for molecular function prediction in inadequately annotated species from understudied taxonomic kingdoms.
| S-EPMC8647222 | biostudies-literature
Project description:The results of our recent study on mathematical modeling of microbial genome evolution indicate that, on average, genomes of bacteria and archaea evolve in the regime of mutation-selection balance defined by positive selection coefficients associated with gene acquisition that is counter-acted by the intrinsic deletion bias. This analysis was based on the strong assumption that parameters of genome evolution are universal across the diversity of bacteria and archaea, and yielded extremely low values of the selection coefficient. Here we further refine the modeling approach by taking into account evolutionary factors specific for individual groups of microbes using two independent fitting strategies, an ad hoc hard fitting scheme and a mixture model. The resulting estimate of the mean selection coefficient of s∼10-10 associated with the gain of one gene implies that, on average, acquisition of a gene is beneficial, and that microbial genomes typically evolve under a weak selection regime that might transition to strong selection in highly abundant organisms with large effective population sizes. The apparent selective pressure towards larger genomes is balanced by the deletion bias, which is estimated to be consistently greater than unity for all analyzed groups of microbes. The estimated values of s are more realistic than the lower values obtained previously, indicating that global and group-specific evolutionary factors synergistically affect microbial genome evolution that seems to be driven primarily by adaptation to existence in diverse niches.
| S-EPMC5898727 | biostudies-literature
Project description:Homologs of the eukaryotic DNA-end-binding protein Ku were identified in several bacterial and one archeal genome using iterative database searches with sequence profiles. Identification of prokaryotic Ku homologs allowed the dissection of the Ku protein sequences into three distinct domains, the Ku core that is conserved in eukaryotes and prokaryotes, a derived von Willebrand A domain that is fused to the amino terminus of the core in eukaryotic Ku proteins, and the newly recognized helix-extension-helix (HEH) domain that is fused to the carboxyl terminus of the core in eukaryotes and in one of the Ku homologs from the Actinomycete Streptomyces coelicolor. The version of the HEH domain present in eukaryotic Ku proteins represents the previously described DNA-binding domain called SAP. The Ku homolog from S. coelicolor contains a distinct version of the HEH domain that belongs to a previously unnoticed family of nucleic-acid-binding domains, which also includes HEH domains from the bacterial transcription termination factor Rho, bacterial and eukaryotic lysyl-tRNA synthetases, bacteriophage T4 endonuclease VII, and several uncharacterized proteins. The distribution of the Ku homologs in bacteria coincides with that of the archeal-eukaryotic-type DNA primase and genes for prokaryotic Ku homologs form predicted operons with genes coding for an ATP-dependent DNA ligase and/or archeal-eukaryotic-type DNA primase. Some of these operons additionally encode an uncharacterized protein that may function as nuclease or an Slx1p-like predicted nuclease containing a URI domain. A hypothesis is proposed that the Ku homolog, together with the associated gene products, comprise a previously unrecognized prokaryotic system for repair of double-strand breaks in DNA.
| S-EPMC311082 | biostudies-literature
Project description:The cell cycle is a highly conserved, continuous process which controls faithful replication and division of cells. Single-cell technologies have enabled increasingly precise measurements of the cell cycle as both as a biological process of interest and as a possible confounding factor. Despite its importance and conservation, there is no universally applicable approach to infer position in the cell cycle with high-resolution from single-cell RNA-seq data. Here, we present tricycle, an R/Bioconductor package,to address this challenge by leveraging key features of the biology of the cell cycle, the mathematical properties of principal component analysis of periodic functions, and the ubiquitous applicability of transfer learning. We show that tricycle can predict any cell’s position in the cell cycle regardless of the cell type, species of origin, and even sequencing assay. The accuracy of tricycle compares favorably to gold-standard experimental assays which generally require specialized measurements in specifically constructed in vitro systems. Unlike gold-standard assays, tricycle is applicable to any single-cell RNA-seq dataset. Tricycle is highly scalable, universally accurate, and eminently pertinent for atlas-level data.
2021-04-07 | GSE171636 | GEO