Evidence classification of high-throughput protocols and confidence integration in RegulonDB

Flexible gold standards for transcription factor regulatory interactions in Escherichia coli K-12: architecture of evidence types

Post-genomic implementations have expanded the experimental strategies to identify elements involved in the regulation of transcription initiation. Here, we present for the first time a detailed analysis of the sources of knowledge supporting the collection of transcriptional regulatory interactions (RIs) of Escherichia coli K-12. An RI groups the transcription factor, its effect (positive or negative) and the regulated target, a promoter, a gene or transcription unit. We improved the evidence codes so that specific methods are incorporated and classified into independent groups. On this basis we updated the computation of confidence levels, weak, strong, or confirmed, for the collection of RIs. These updates enabled us to map the RI set to the current collection of HT TF-binding datasets from ChIP-seq, ChIP-exo, gSELEX and DAP-seq in RegulonDB, enriching in this way the evidence of close to one-quarter (1329) of RIs from the current total 5446 RIs. Based on the new computational capabilities of our improved annotation of evidence sources, we can now analyze the internal architecture of evidence, their categories (experimental, classical, HT, computational), and confidence levels. This is how we know that the joint contribution of HT and computational methods increase the overall fraction of reliable RIs (the sum of confirmed and strong evidence) from 49% to 71%. Thus, the current collection has 3912 reliable RIs, with 2718 or 70% of them with classical evidence which can be used to benchmark novel HT methods. Users can selectively exclude the method they want to benchmark, or keep for instance only the confirmed interactions. The recovery of regulatory sites in RegulonDB by the different HT methods ranges between 33% by ChIP-exo to 76% by ChIP-seq although as discussed, many potential confounding factors limit their interpretation. The collection of improvements reported here provides a solid foundation to incorporate new methods and data, and to further integrate the diverse sources of knowledge of the different components of the transcriptional regulatory network. There is no other genomic database that offers this comprehensive high-quality architecture of knowledge supporting a corpus of transcriptional regulatory interactions.

RegulonDB v12.0: a comprehensive resource of transcriptional regulation in E. coli K-12

RegulonDB is a database that contains the most comprehensive corpus of knowledge of the regulation of transcription initiation of Escherichia coli K-12, including data from both classical molecular biology and high-throughput methodologies. Here, we describe biological advances since our last NAR paper of 2019. We explain the changes to satisfy FAIR requirements. We also present a full reconstruction of the RegulonDB computational infrastructure, which has significantly improved data storage, retrieval and accessibility and thus supports a more intuitive and user-friendly experience. The integration of graphical tools provides clear visual representations of genetic regulation data, facilitating data interpretation and knowledge integration. RegulonDB version 12.0 can be accessed at https://regulondb.ccg.unam.mx.

A Gold Standard for Transcription Factor Regulatory Interactions in Escherichia coli K-12: Architecture of Evidence Types

Post-genomic implementations have expanded the experimental strategies to identify elements involved in the regulation of transcription initiation. As new methodologies emerge, a natural step is to compare their results with those from established methodologies, such as the classic methods of molecular biology used to characterize transcription factor binding sites, promoters, or transcription units. In the case of Escherichia coli K-12, the best-studied microorganism, for the last 30 years we have continuously gathered such knowledge from original scientific publications, and have organized it in two databases, RegulonDB and EcoCyc. Furthermore, since RegulonDB version 11.0 (1), we offer comprehensive datasets of binding sites from chromatin immunoprecipitation combined with sequencing (ChIP-seq), ChIP combined with exonuclease digestion and next-generation sequencing (ChIP-exo), genomic SELEX screening (gSELEX), and DNA affinity purification sequencing (DAP-seq) HT technologies, as well as additional datasets for transcription start sites, transcription units and RNA sequencing (RNA-seq) expression profiles. Here, we present for the first time an analysis of the sources of knowledge supporting the collection of transcriptional regulatory interactions (RIs) of E. coli K-12. An RI is formed by the transcription factor, its positive or negative effect on a promoter, a gene or transcription unit. We improved the evidence codes so that the specific methods are described, and we classified them into seven independent groups. This is the basis for our updated computation of confidence levels, weak, strong, or confirmed, for the collection of RIs. We compare the confidence levels of the RI collection before and after adding HT evidence illustrating how knowledge will change as more HT data and methods appear in the future. Users can generate subsets filtering out the method they want to benchmark and avoid circularity, or keep for instance only the confirmed interactions. The comparison of different HT methods with the available datasets indicate that ChIP-seq recovers the highest fraction (>70%) of binding sites present in RegulonDB followed by gSELEX, DAP-seq and ChIP-exo. There is no other genomic database that offers this comprehensive high-quality anatomy of evidence supporting a corpus of transcriptional regulatory interactions.

Genome-scale prediction of bacterial promoters

A key step in the transcription of RNA is the binding of the RNA polymerase protein complex to a short promoter sequence that is typically upstream of the gene to be expressed. Automated identification of promoters would serve as a valuable complement to experimental validation in determining which genes are likely to be expressed and when; however, promoter sequences are short and highly variable, which makes them very difficult to accurately classify. The many tools developed to identify promoters in DNA have generally been tested on small and balanced subsets of genomic sequence, and the results may not reflect their expected performance on genomes with millions of DNA base pairs where promoters are likely to comprise less than ∼1% of the sequence. Here we introduce Expositor, a neural-network-based method that uses different types of DNA encodings and tunable sensitivity and specificity parameters. Expositor showed higher sensitivity and precision on the E. coli K-12 MG1655 chromosome than other tested approaches. Expositor predictions were more consistent in the homologous subset of sequence from a strain of Salmonella than they were with another strain of E. coli. We also examined the accuracy of Expositor in distinguishing different classes of promoters and found that misclassification between classes was consistent with the biological similarity between promoters.

Curation, inference, and assessment of a globally reconstructed gene regulatory network for Streptomyces coelicolor

Streptomyces coelicolor A3(2) is a model microorganism for the study of Streptomycetes, antibiotic production, and secondary metabolism in general. Even though S. coelicolor has an outstanding variety of regulators among bacteria, little effort to globally study its transcription has been made. We manually curated 29 years of literature and databases to assemble a meta-curated experimentally-validated gene regulatory network (GRN) with 5386 genes and 9707 regulatory interactions (~ 41% of the total expected interactions). This provides the most extensive and up-to-date reconstruction available for the regulatory circuitry of this organism. Only ~ 6% (534/9707) are supported by experiments confirming the binding of the transcription factor to the upstream region of the target gene, the so-called “strong” evidence. While for the remaining interactions there is no confirmation of direct binding. To tackle network incompleteness, we performed network inference using several methods (including two proposed here) for motif identification in DNA sequences and GRN inference from transcriptomics. Further, we contrasted the structural properties and functional architecture of the networks to assess the reliability of the predictions, finding the inference from DNA sequence data to be the most trustworthy approach. Finally, we show two applications of the inferred and the curated networks. The inference allowed us to propose novel transcription factors for the key Streptomyces antibiotic regulatory proteins (SARPs). The curated network allowed us to study the conservation of the system-level components between S. coelicolor and Corynebacterium glutamicum. There we identified the basal machinery as the common signature between the two organisms. The curated networks were deposited in Abasy Atlas (https://abasy.ccg.unam.mx/) while the inferences are available as Supplementary Material.

Abasy Atlas v2.2: The most comprehensive and up-to-date inventory of meta-curated, historical, bacterial regulatory networks, their completeness and system-level characterization

Some organism-specific databases about regulation in bacteria have become larger, accelerated by high-throughput methodologies, while others are no longer updated or accessible. Each database homogenize its datasets, giving rise to heterogeneity across databases. Such heterogeneity mainly encompasses different names for a gene and different network representations, generating duplicated interactions that could bias network analyses. Abasy (Across-bacteria systems) Atlas consolidates information from different sources into meta-curated regulatory networks in bacteria. The high-quality networks in Abasy Atlas enable cross-organisms analyses, such as benchmarking studies where gold standards are required. Nevertheless, network incompleteness still casts doubts on the conclusions of network analyses, and available sampling methods cannot reflect the curation process. To tackle this problem, the updated version of Abasy Atlas presented in this work provides historical snapshots of regulatory networks. Thus, network analyses can be performed at different completeness levels, making possible to identify potential bias and to predict future results. We leverage the recently found constraint in the complexity of regulatory networks to develop a novel model to quantify the total number of regulatory interactions as a function of the genome size. This completeness estimation is a valuable insight that may aid in the daunting task of network curation, prediction, and validation. The new version of Abasy Atlas provides 76 networks (204,282 regulatory interactions) covering 42 bacteria (64% Gram-positive and 36% Gram-negative) distributed in 9 species (Mycobacterium tuberculosis, Bacillus subtilis, Escherichia coli, Corynebacterium glutamicum, Staphylococcus aureus, Pseudomonas aeruginosa, Streptococcus pyogenes, Streptococcus pneumoniae, and Streptomyces coelicolor), containing 8459 regulons and 4335 modules.

Database URL: https://abasy.ccg.unam.mx/.

RegulonDB v 10.5: tackling challenges to unify classic and high throughput knowledge of gene regulation in E. coli K-12

RegulonDB, first published 20 years ago, is a comprehensive electronic resource about regulation of transcription initiation of Escherichia coli K-12 with decades of knowledge from classic molecular biology experiments, and recently also from high-throughput genomic methodologies. We curated the literature to keep RegulonDB up to date, and initiated curation of ChIP and gSELEX experiments. We estimate that current knowledge describes between 10% and 30% of the expected total number of transcription factor- gene regulatory interactions in E. coli. RegulonDB provides datasets for interactions for which there is no evidence that they affect expression, as well as expression datasets. We developed a proof of concept pipeline to merge binding and expression evidence to identify regulatory interactions. These datasets can be visualized in the RegulonDB JBrowse. We developed the Microbial Conditions Ontology with a controlled vocabulary for the minimal properties to reproduce an experiment, which contributes to integrate data from high throughput and classic literature. At a higher level of integration, we report Genetic Sensory-Response Units for 200 transcription factors, including their regulation at the metabolic level, and include summaries for 70 of them. Finally, we summarize our research with Natural language processing strategies to enhance our biocuration work.

Limits to a classic paradigm: most transcription factors in E. coli regulate genes involved in multiple biological processes

Transcription factors (TFs) are important drivers of cellular decision-making. When bacteria encounter a change in the environment, TFs alter the expression of a defined set of genes in order to adequately respond. It is commonly assumed that genes regulated by the same TF are involved in the same biological process. Examples of this are methods that rely on coregulation to infer function of not-yet-annotated genes. We have previously shown that only 21% of TFs involved in metabolism regulate functionally homogeneous genes, based on the proximity of the gene products’ catalyzed reactions in the metabolic network. Here, we provide more evidence to support the claim that a 1-TF/1-process relationship is not a general property. We show that the observed functional heterogeneity of regulons is not a result of the quality of the annotation of regulatory interactions, nor the absence of protein–metabolite interactions, and that it is also present when function is defined by Gene Ontology terms. Furthermore, the observed functional heterogeneity is different from the one expected by chance, supporting the notion that it is a biological property. To further explore the relationship between transcriptional regulation and metabolism, we analyzed five other types of regulatory groups and identified complex regulons (i.e. genes regulated by the same combination of TFs) as the most functionally homogeneous, and this is supported by coexpression data. Whether higher levels of related functions exist beyond metabolism and current functional annotations remains an open question.

Using RegulonDB, the Escherichia coli K-12 Gene Regulatory Transcriptional Network Database

In RegulonDB, for over 25 years, we have been gathering knowledge by manual curation from original scientific literature on the regulation of transcription initiation and genome organization in transcription units of the Escherichia coli K-12 genome. This unit describes six basic protocols that can serve as a guiding introduction to the main content of the current version (v9.4) of this electronic resource. These protocols include general navigation as well as searching for specific objects such as genes, gene products, transcription units, promoters, transcription factors, coexpression, and genetic sensory response units or GENSOR Units. In these protocols, the user will find an initial introduction to the concepts pertinent to the protocol, the content obtained when performing the given navigation, and the necessary resources for carrying out the protocol. This easy-to-follow presentation should help anyone interested in quickly seeing all that is currently offered in RegulonDB, including position weight matrices of transcription factors, coexpression values based on published microarrays, and the GENSOR Units unique to RegulonDB that offer regulatory mechanisms in the context of their signals and metabolic consequences. © 2018 by John Wiley & Sons, Inc.

Evolutionary constraints on the complexity of genetic regulatory networks allow predictions of the total number of genetic interactions

Genetic regulatory networks (GRNs) have been widely studied, yet there is a lack of understanding with regards to the final size and properties of these networks, mainly due to no network currently being complete. In this study, we analyzed the distribution of GRN structural properties across a large set of distinct prokaryotic organisms and found a set of constrained characteristics such as network density and number of regulators. Our results allowed us to estimate the number of interactions that complete networks would have, a valuable insight that could aid in the daunting task of network curation, prediction, and validation. Using state-of-the-art statistical approaches, we also provided new evidence to settle a previously stated controversy that raised the possibility of complete biological networks being random and therefore attributing the observed scale-free properties to an artifact emerging from the sampling process during network discovery. Furthermore, we identified a set of properties that enabled us to assess the consistency of the connectivity distribution for various GRNs against different alternative statistical distributions. Our results favor the hypothesis that highly connected nodes (hubs) are not a consequence of network incompleteness. Finally, an interaction coverage computed for the GRNs as a proxy for completeness revealed that high-throughput based reconstructions of GRNs could yield biased networks with a low average clustering coefficient, showing that classical targeted discovery of interactions is still needed.

A unified resource for transcriptional regulation in Escherichia coli K-12 incorporating high-throughput-generated binding data into RegulonDB version 10.0

BACKGROUND

Our understanding of the regulation of gene expression has benefited from the availability of high-throughput technologies that interrogate the whole genome for the binding of specific transcription factors and gene expression profiles. In the case of widely used model organisms, such as Escherichia coli K-12, the new knowledge gained from these approaches needs to be integrated with the legacy of accumulated knowledge from genetic and molecular biology experiments conducted in the pre-genomic era in order to attain the deepest level of understanding possible based on the available data.

RESULTS

In this paper, we describe an expansion of RegulonDB, the database containing the rich legacy of decades of classic molecular biology experiments supporting what we know about gene regulation and operon organization in E. coli K-12, to include the genome-wide dataset collections from 32 ChIP and 19 gSELEX publications, in addition to around 60 genome-wide expression profiles relevant to the functional significance of these datasets and used in their curation. Three essential features for the integration of this information coming from different methodological approaches are: first, a controlled vocabulary within an ontology for precisely defining growth conditions; second, the criteria to separate elements with enough evidence to consider them involved in gene regulation from isolated transcription factor binding sites without such support; and third, an expanded computational model supporting this knowledge. Altogether, this constitutes the basis for adequately gathering and enabling the comparisons and integration needed to manage and access such wealth of knowledge.

CONCLUSIONS

This version 10.0 of RegulonDB is a first step toward what should become the unifying access point for current and future knowledge on gene regulation in E. coli K-12. Furthermore, this model platform and associated methodologies and criteria can be emulated for gathering knowledge on other microbial organisms.

Global transcriptional regulatory network for Escherichia coli robustly connects gene expression to transcription factor activities

Transcriptional regulatory networks (TRNs) have been studied intensely for >25 y. Yet, even for the Escherichia coli TRN-probably the best characterized TRN-several questions remain. Here, we address three questions: (i) How complete is our knowledge of the E. coli TRN; (ii) how well can we predict gene expression using this TRN; and (iii) how robust is our understanding of the TRN? First, we reconstructed a high-confidence TRN (hiTRN) consisting of 147 transcription factors (TFs) regulating 1,538 transcription units (TUs) encoding 1,764 genes. The 3,797 high-confidence regulatory interactions were collected from published, validated chromatin immunoprecipitation (ChIP) data and RegulonDB. For 21 different TF knockouts, up to 63% of the differentially expressed genes in the hiTRN were traced to the knocked-out TF through regulatory cascades. Second, we trained supervised machine learning algorithms to predict the expression of 1,364 TUs given TF activities using 441 samples. The algorithms accurately predicted condition-specific expression for 86% (1,174 of 1,364) of the TUs, while 193 TUs (14%) were predicted better than random TRNs. Third, we identified 10 regulatory modules whose definitions were robust against changes to the TRN or expression compendium. Using surrogate variable analysis, we also identified three unmodeled factors that systematically influenced gene expression. Our computational workflow comprehensively characterizes the predictive capabilities and systems-level functions of an organism's TRN from disparate data types.

Global transcriptional start site mapping in Geobacter sulfurreducens during growth with two different electron acceptors

Geobacter sulfurreducens is an anaerobic soil bacterium that is involved in biogeochemical cycles of elements such as Fe and Mn. Although significant progress has been made in the understanding of the electron transfer processes in G. sulfurreducens, little is known about the regulatory mechanisms involved in their control. To expand the study of gene regulation in G. sulfurreducens, we carried out a genome-wide identification of transcription start sites (TSS) by 5'RACE and by deep RNA sequencing of primary mRNAs in two growth conditions. TSSs were identified along G. sulfurreducens genome and over 50% of them were located in the upstream region of the associated gene, and in some cases we detected genes with more than one TSS. Our global mapping of TSSs contributes with valuable information, which is needed for the study of transcript structure and transcription regulation signals and can ultimately contribute to the understanding of transcription initiation phenomena in G. sulfurreducens.

RegulonDB version 9.0: high-level integration of gene regulation, coexpression, motif clustering and beyond

RegulonDB (http://regulondb.ccg.unam.mx) is one of the most useful and important resources on bacterial gene regulation,as it integrates the scattered scientific knowledge of the best-characterized organism, Escherichia coli K-12, in a database that organizes large amounts of data. Its electronic format enables researchers to compare their results with the legacy of previous knowledge and supports bioinformatics tools and model building. Here, we summarize our progress with RegulonDB since our last Nucleic Acids Research publication describing RegulonDB, in 2013. In addition to maintaining curation up-to-date, we report a collection of 232 interactions with small RNAs affecting 192 genes, and the complete repertoire of 189 Elementary Genetic Sensory-Response units (GENSOR units), integrating the signal, regulatory interactions, and metabolic pathways they govern. These additions represent major progress to a higher level of understanding of regulated processes. We have updated the computationally predicted transcription factors, which total 304 (184 with experimental evidence and 120 from computational predictions); we updated our position-weight matrices and have included tools for clustering them in evolutionary families. We describe our semiautomatic strategy to accelerate curation, including datasets from high-throughput experiments, a novel coexpression distance to search for ‘neighborhood’ genes to known operons and regulons, and computational developments.

Assisted curation of regulatory interactions and growth conditions of OxyR in E. coli K-12

Given the current explosion of data within original publications generated in the field of genomics, a recognized bottleneck is the transfer of such knowledge into comprehensive databases. We have for years organized knowledge on transcriptional regulation reported in the original literature of Escherichia coli K-12 into RegulonDB (http://regulondb.ccg.unam.mx), our database that is currently supported by >5000 papers. Here, we report a first step towards the automatic biocuration of growth conditions in this corpus. Using the OntoGene text-mining system (http://www.ontogene.org), we extracted and manually validated regulatory interactions and growth conditions in a new approach based on filters that enable the curator to select informative sentences from preprocessed full papers. Based on a set of 48 papers dealing with oxidative stress by OxyR, we were able to retrieve 100% of the OxyR regulatory interactions present in RegulonDB, including the transcription factors and their effect on target genes. Our strategy was designed to extract, as we did, their growth conditions. This result provides a proof of concept for a more direct and efficient curation process, and enables us to define the strategy of the subsequent steps to be implemented for a semi-automatic curation of original literature dealing with regulation of gene expression in bacteria. This project will enhance the efficiency and quality of the curation of knowledge present in the literature of gene regulation, and contribute to a significant increase in the encoding of the regulatory network of E. coli.

RegulonDB Database URL:http://regulondb.ccg.unam.mx

OntoGene URL:http://www.ontogene.org

RegulonDB v8.0: omics data sets, evolutionary conservation, regulatory phrases, cross-validated gold standards and more

This article summarizes our progress with RegulonDB (http://regulondb.ccg.unam.mx/) during the past 2 years. We have kept up-to-date the knowledge from the published literature regarding transcriptional regulation in Escherichia coli K-12. We have maintained and expanded our curation efforts to improve the breadth and quality of the encoded experimental knowledge, and we have implemented criteria for the quality of our computational predictions. Regulatory phrases now provide high-level descriptions of regulatory regions. We expanded the assignment of quality to various sources of evidence, particularly for knowledge generated through high-throughput (HT) technology. Based on our analysis of most relevant methods, we defined rules for determining the quality of evidence when multiple independent sources support an entry. With this latest release of RegulonDB, we present a new highly reliable larger collection of transcription start sites, a result of our experimental HT genome-wide efforts. These improvements, together with several novel enhancements (the tracks display, uploading format and curational guidelines), address the challenges of incorporating HT-generated knowledge into RegulonDB. Information on the evolutionary conservation of regulatory elements is also available now. Altogether, RegulonDB version 8.0 is a much better home for integrating knowledge on gene regulation from the sources of information currently available.