Biology subjects

Deutsch, E. W.

Publications and source records attributed to Deutsch, E. W..

At least 19 recordsLinked to original sources

bioRxiv2025

The E. coli PeptideAtlas Build: Characterizing the observed Escherichia coli pan-proteome and its post-translational modifications

Escherichia coli is a widely used model organism in molecular biology. Despite its pivotal role, a comprehensive proteome resource covering the E. coli pan-proteome and its post-translational modifications (PTMs) has been lacking. Here we present the E. coli PeptideAtlas build, the first comprehensive pan-proteome analysis of E. coli, generated from 40 high-quality public and in-house datasets spanning a broad diversity of strains, sample types, and experimental conditions, and comprising over 73 million MS/MS spectra. All datasets were reprocessed using both a closed search (Trans-Proteomic Pipeline using MSFragger) and an open search (ionbot). The E. coli PeptideAtlas build provides evidence for 4,755 proteins, including 1,410 previously lacking protein-level support in UniProt. The resource offers protein coverage, modification sites, raw spectra with matched peptides, and manually annotated metadata for the E. coli pan proteome. PTM profiling identified over 10,000 modification sites, including phosphorylation (3,806), acetylation (754), methylation (730), glutathionylation (352) and phosphoribosylation (226). Analysis of the glutathionylation sites revealed potential links to metal binding regulation. We also detected proteins likely associated with phages, underscoring the value of pan-proteomic approaches for studying host-phage interactions. All identifications are publicly accessible and traceable through the PeptideAtlas interface. We expect that the E. coli PeptideAtlas build will provide a useful resource for the community, which supports, for example, targeted MS experiment design, PTM enrichment method development, and strain typing. It allows straightforward lookups of protein and peptide identifications and facilitates comparative proteomic analyses by enabling the assessment of protein presence and variability across different E. coli strains. The build is available at https://db.systemsbiology.net/sbeams/cgi/PeptideAtlas/buildDetails?atlas_build_id=585.

bioinformatics↗

bioRxiv2025

An expanded reference catalog of translated open reading frames for biomedical research

Non-canonical (i.e., unannotated) open reading frames (ncORFs) have until recently been omitted from reference genome annotations, despite evidence of their translation, limiting their incorporation into biomedical research. To address this, in 2022, we initiated the TransCODE consortium and built the first community-driven consensus catalog of human ncORFs, which was openly distributed to the research community via Ensembl-GENCODE. While this catalog represented a starting point for reference ncORF annotation, major technical and scientific issues remained. In particular, this initial catalogue had no standardized framework to judge the evidence of translation for individual ncORFs. Here, we present an expanded and refined catalog of the human reference annotation of ncORFs. By incorporating more datasets and by lifting constraints on ORF length and start-codon, we define a comprehensive set of 28,359 ncORFs that is nearly four times the size of the previous catalog. Furthermore, to aid users who wish to work with ncORFs with the strongest and most reproducible signals of translation, we utilized a data-driven framework (i.e. translation signature scores) to assess the accumulated evidence for any individual ncORF. Using this approach, we derive a subset of 7,888 ncORFs with translation evidence on par with canonical protein-coding genes, which we refer to as the Primary set. This set can serve as a reliable reference for downstream analyses and validation, with a particular emphasis on high quality. Overall, this update reflects continual community-driven efforts to make ncORFs accessible and actionable to the broader research public and further iterations of the catalog will continue to expand and refine this resource.

genomics↗

bioRxiv2025

PloverDB: A high-performance platform for serving biomedical knowledge graphs as standards-compliant web APIs

Knowledge graphs are increasingly being used to integrate heterogeneous biomedical knowledge and data. General-purpose graph database management systems such as Neo4j are often used to host and search knowledge graphs, but such tools come with overhead and leave biomedical-specific standards compliance and reasoning to the user. Interoperability across biomedical knowledge bases and reasoning systems necessitates the use of standards such as those adopted by the Biomedical Data Translator consortium. We present PloverDB, a comprehensive software platform for hosting and efficiently serving biomedical knowledge graphs as standards-compliant web application programming interfaces. In addition to fundamental back-end knowledge reasoning tasks, PloverDB automatically handles entity resolution, exposure of standardized metadata and test data, and multiplexing of knowledge graphs, all in a single platform designed specifically for efficient query answering and ease of deployment. PloverDB increases data accessibility and utility by allowing data providers to quickly serve their biomedical knowledge graphs as standards-compliant web services. Availability and Implementation: PloverDBs source code and technical documentation are publicly available under an MIT License at github:RTXteam/PloverDB, archived on Zenodo at doi:10.5281/zenodo.15454600. Contactamykglen@gmail.com

bioinformatics↗

bioRxiv2025

A pan-gene catalogue of Asian cultivated rice

The rice genome underpins fundamental research and breeding, but the Nipponbare (japonica) reference does not fully encompass the genetic diversity of Asian rice. To address this gap, the Rice Population Reference Panel (RPRP) was developed, comprising high-quality assemblies of 16 rice cultivars to represent japonica, indica, aus, and aromatic varietal groups. The RPRP has been consistently annotated, supported by extensive experimental data and here we report the computational assignment, characterization and dissemination of stably identified pan-genes. We identified 25,178 core pan-genes shared across all cultivars, alongside cultivar-specific and family-enriched genes. Core genes exhibit higher gene expression and proteomic evidence, higher confidence protein domains and AlphaFold structures, while cultivar-specific genes were enriched for domains under selective breeding pressure, such as for disease resistance. This resource, integrated into public databases, enables researchers to explore genetic and functional diversity via a population-aware "reference guide" across rice genomes, advancing both basic and applied research.

genomics↗

bioRxiv2025

Detection of human unannotated microproteins by mass spectrometry-based proteomics: a community assessment

Thousands of short open reading frames (sORFs) are translated outside of annotated coding sequences. Recent studies have pioneered searching for sORF-encoded microproteins in mass spectrometry (MS)- based proteomics and peptidomics datasets. Here, we assessed literature-reported MS-based identifications of unannotated human proteins. We find that studies vary by three orders of magnitude in the number of unannotated proteins they report. Of nearly 10,000 reported sORF-encoded peptides, 96% were unique to a single study, and 12% mapped to annotated proteins or proteoforms. Manual curation of a benchmark dataset of 406 manually evaluated spectra from 204 sORF-encoded proteins revealed large variation in peptide-spectrum match (PSM) quality between studies, with immunopeptidomics studies generally reporting higher quality PSMs than conventional enzymatic digests of whole cell lysates. We estimate that 65% of predicted sORF-encoded protein detections in immunopeptidomics studies were supported by high-quality PSMs versus 7.8% in non-immunopeptidomics datasets. Our work stresses the need for standardized protocols and analysis workflows to guide future advancements in microprotein detection by MS towards uncovering how many human microproteins exist.

genomics↗

bioRxiv2025

Applying a conservation-based approach for predicting novel phosphorylation sites in eukaryotes and evaluating their functional relevance

Protein phosphorylation, a key post-translational modification, is central to cellular signalling and disease pathogenesis. The development of high-throughput proteomics pipelines has led to the discovery of large numbers of phosphorylated protein motifs and sites (phosphosites) across many eukaryotic species. However, the majority of phosphosites are reported from human cellular sources, with most species having only a few experimentally confirmed or computationally predicted phosphosites. Furthermore, only a small fraction of the characterised human phosphoproteome has an annotated functional role. A common method of predicting functionally relevant phosphosites is conservation analysis. However, extensive evolutionary studies involving large numbers of species are scarce. In this study, we explore the conservation of 20,751 confident human phosphosites across 100 eukaryotic species. We link the observed conservation patterns to the functional relevance of phosphosites and investigate the evolution of associated protein domains and kinases. We identify several protein functions that are likely regulated by phosphorylation, including ancient functions conserved across all life and relatively new functions only conserved in species closely related to humans. We demonstrate the importance of conservation analysis in identifying organisms suitable as biological models for studying conserved signalling pathways relevant to human biology and disease. Finally, we use human protein sequences as a reference to apply conservation analysis and predict over 1,000,000 potential phosphosites in analysed eukaryotes. Our results can ultimately be used to improve proteome annotations of several species and direct downstream research surrounding the evolution and functional relevance of eukaryotic phosphosites.

bioinformatics↗

bioRxiv2024

Candida albicans a comprehensive view of the proteome

We describe a new release of the Candida albicans PeptideAtlas proteomics spectral resource (build 2024-03), providing a sequence coverage of 79.5% at the canonical protein level, matched mass spectrometry spectra, and experimental evidence identifying 3382 and 536 phosphorylated serine and threonine sites with false localization rates of 1% and 5.3%, respectively. We provide a tutorial on how to use the PeptideAtlas and associated tools to access this information. The C. albicans PeptideAtlas summary web page provides "Build overview", "PTM coverage", "Experiment contribution", and "Dataset contribution" information. The protein and peptide information can also be accessed via the Candida Genome Database via hyperlinks on each protein page. This allows users to peruse identified peptides, protein coverage, post-translational modifications (PTMs), and experiments identifying each protein. Given the value of understanding the PTM landscape in the sequence of each protein, a more detailed explanation of how to interpret and analyse PTM results is provided in the PeptideAtlas of this important pathogen. Candida albicans PeptideAtlas web page: https://db.systemsbiology.net/sbeams/cgi/PeptideAtlas/buildDetails?atlas_build_id=578

microbiology↗

bioRxiv2024

Searching for Sulfotyrosines (sY) in a HA(pY)STACK

Protein sulfation can be crucial in regulating protein-protein interactions but remains largely underexplored. Sulfation is near-isobaric to phosphorylation, making it particularly challenging to investigate using mass spectrometry. The degree to which tyrosine sulfation (sY) is misidentified as phosphorylation (pY) is thus an unresolved concern. This study explores the extent of sY misidentification within the human phosphoproteome by distinguishing between sulfation and phosphorylation based on their mass difference. Using Gaussian mixture models (GMMs), we screened [~]45M peptide-spectrum matches (PSMs) from the PeptideAtlas Human Phosphoproteome build for peptidoforms with mass error shifts indicative of sulfation. This analysis pinpointed 104 candidate sulfated peptidoforms, backed-up by Gene Ontology (GO) terms and custom terms linked to sulfation. False positive filtering by manual annotation resulted in 31 convincing peptidoforms spanning 7 known and 7 novel sY sites. Y47 in Calumenin was particularly intriguing since mass error shifts, acidic motif conservation, and MS2 neutral loss patterns characteristic of sulfation provided strong evidence that this site is sulfated rather than phosphorylated. Overall, although misidentification of sulfation in phosphoproteomics datasets derived from cell and tissue intracellular extracts can occur, it appears relatively rare and should not be considered a substantive confounding factor in high-quality phosphoproteomics datasets.

bioinformatics↗

bioRxiv2024

High-quality peptide evidence for annotating non-canonical open reading frames as human proteins

A major scientific drive is to characterize the protein-coding genome as it provides the primary basis for the study of human health. But the fundamental question remains: what has been missed in prior genomic analyses? Over the past decade, the translation of non-canonical open reading frames (ncORFs) has been observed across human cell types and disease states, with major implications for proteomics, genomics, and clinical science. However, the impact of ncORFs has been limited by the absence of a large-scale understanding of their contribution to the human proteome. Here, we report the collaborative efforts of stakeholders in proteomics, immunopeptidomics, Ribo-seq ORF discovery, and gene annotation, to produce a consensus landscape of protein-level evidence for ncORFs. We show that at least 25% of a set of 7,264 ncORFs give rise to translated gene products, yielding over 3,000 peptides in a pan-proteome analysis encompassing 3.8 billion mass spectra from 95,520 experiments. With these data, we developed an annotation framework for ncORFs and created public tools for researchers through GENCODE and PeptideAtlas. This work will provide a platform to advance ncORF-derived proteins in biomedical discovery and, beyond humans, diverse animals and plants where ncORFs are similarly observed.

molecular biology↗

bioRxiv2023

The Zea mays PeptideAtlas; a new maize community resource

We developed the Maize PeptideAtlas resource (www.peptideatlas.org/builds/maize) to help solve questions about the maize proteome. Publicly available raw tandem mass spectrometry (MS/MS) data for maize were collected from ProteomeXchange and reanalyzed through a uniform processing and metadata annotation pipeline. These data are from a wide range of genetic backgrounds, including the inbred lines B73 and W22, many hybrids and their respective parents. Samples were collected from field trials, controlled environmental conditions, a range of (a)biotic conditions and different tissues, cell types and subcellular fractions. The protein search space included different maize genome annotations for the B73 inbred line from MaizeGDB, UniProtKB, NCBI RefSeq and for the W22 inbred line. 445 million MS/MS spectra were searched, of which 120 million were matched to 0.37 million distinct peptides. Peptides were matched to 66.2% of the proteins (one isoform per protein coding gene) in the most recent B73 nuclear genome annotation (v5). Furthermore, most conserved plastid- and mitochondrial-encoded proteins (NCBI RefSeq annotations) were identified. Peptides and proteins identified in the other searched B73 genome annotations will aid to improve maize genome annotation. We also illustrate high confidence detection of unique W22 proteins. N-terminal acetylation, phosphorylation, ubiquitination, and three lysine acylations (K-acetyl, K-malonyl, K-hydroxyisobutyryl) were identified and can be inspected through a PTM viewer in PeptideAtlas. All matched MS/MS-derived peptide data are linked to spectral, technical and biological metadata. This new PeptideAtlas is integrated with community resources including MaizeGDB at https://www.maizegdb.org/ and a peptide track in JBrowse. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=119 SRC="FIGDIR/small/572651v2_ufig1.gif" ALT="Figure 1"> View larger version (28K): org.highwire.dtl.DTLVardef@1ec9005org.highwire.dtl.DTLVardef@1e35f34org.highwire.dtl.DTLVardef@7f5b2corg.highwire.dtl.DTLVardef@13acefd_HPS_FORMAT_FIGEXP M_FIG C_FIG

plant biology↗

bioRxiv2023

A meta-analysis of rice phosphoproteomics data to understand variation in cell signalling across the rice pan-genome

Phosphorylation is the most studied post-translational modification, and has multiple biological functions. In this study, we have re-analysed publicly available mass spectrometry proteomics datasets enriched for phosphopeptides from Asian rice (Oryza sativa). In total we identified 15,522 phosphosites on serine, threonine and tyrosine residues on rice proteins. We identified sequence motifs for phosphosites, and link motifs to enrichment of different biological processes, indicating different downstream regulation likely caused by different kinase groups. We cross-referenced phosphosites against the rice 3,000 genomes, to identify single amino acid variations (SAAVs) within or proximal to phosphosites that could cause loss of a site in a given rice variety. The data was clustered to identify groups of sites with similar patterns across rice family groups, for example those highly conserved in Japonica, but mostly absent in Aus type rice varieties - known to have different responses to drought. These resources can assist rice researchers to discover alleles with significantly different functional effects across rice varieties. The data has been loaded into UniProt Knowledge-Base - enabling researchers to visualise sites alongside other data on rice proteins e.g. structural models from AlphaFold2, PeptideAtlas and the PRIDE database - enabling visualisation of source evidence, including scores and supporting mass spectra.

bioinformatics↗

bioRxiv2023

Detection and editing of the updated plastid- and mitochondrial-encoded proteomes for Arabidopsis with PeptideAtlas

Arabidopsis thaliana Col-0 has plastid and mitochondrial genomes encoding for over one hundred proteins and several ORFs. Public databases (e.g. Araport11) have redundancy and discrepancies in gene identifiers for these organelle-encoded proteins. RNA editing results in changes to specific amino acid residues or creation of start and stop codons for many of these proteins, but the impact of such RNA editing at the protein level is largely unexplored due to the complexities of detection. This study first assembled the non-redundant set of identifiers, their correct protein sequences, and 452 predicted non-synonymous editing sites of which 56 are edited at lower frequency. Accumulation of edited and/or unedited proteoforms was then determined by searching [~]259 million raw MSMS spectra from ProteomeXchange as part of Arabidopsis PeptideAtlas (www.peptideatlas.org/builds/arabidopsis/). All mitochondrial proteins and all except three plastid-encoded proteins (NDHG/NDH6, PSBM, RPS16), but none of the ORFs, were identified; we suggest that all ORFs and RPS16 are pseudogenes. Detection frequencies for each edit site and type of edit (e.g. S to L/F) were determined at the protein level, cross-referenced against the metadata (e.g. tissue), and evaluated for technical challenges of detection.167 predicted edit sites were detected at the proteome level. Minor frequency sites were indeed also edited at low frequency at the protein level. However, except for sites RPL5-22 and CCB382-124, proteins only accumulate in edited form (>98 -100% edited) even if RNA editing levels are well below 100%. This study establishes that RNA editing for major editing sites is required for stable protein accumulation.

plant biology↗

bioRxiv2023

Mapping the Arabidopsis thaliana proteome in PeptideAtlas and the nature of the unobserved (dark) proteome; strategies towards a complete proteome

This study describes a new release of the Arabidopsis thaliana PeptideAtlas proteomics resource providing protein sequence coverage, matched mass spectrometry (MS) spectra, selected PTMs, and metadata. 70 million MS/MS spectra were matched to the Araport11 annotation, identifying [~]0.6 million unique peptides and 18267 proteins at the highest confidence level and 3396 lower confidence proteins, together representing 78.6% of the predicted proteome. Additional identified proteins not predicted in Araport11 should be considered for building the next Arabidopsis genome annotation. This release identified 5198 phosphorylated proteins, 668 ubiquitinated proteins, 3050 N-terminally acetylated proteins and 864 lysine-acetylated proteins and mapped their PTM sites. MS support was lacking for 21.4% (5896 proteins) of the predicted Araport11 proteome - the dark proteome. This dark proteome is highly enriched for certain (e.g. CLE, CEP, IDA, PSY) but not other (e.g. THIONIN, CAP,) signaling peptides families, E3 ligases, TFs, and other proteins with unfavorable physicochemical properties. A machine learning model trained on RNA expression data and protein properties predicts the probability for proteins to be detected. The model aids in discovery of proteins with short-half life (e.g. SIG1,3 and ERF-VII TFs) and completing the proteome. PeptideAtlas is linked to TAIR, JBrowse, PPDB, SUBA, UniProtKB and Plant PTM Viewer.

plant biology↗

bioRxiv2023

What can Ribo-seq and proteomics tell us about the non-canonical proteome?

Ribosome profiling (Ribo-seq) has proven transformative for our understanding of the human genome and proteome by illuminating thousands of non-canonical sites of ribosome translation outside of the currently annotated coding sequences (CDSs). A conservative estimate suggests that at least 7,000 non-canonical open reading frames (ORFs) are translated, which, at first glance, has the potential to expand the number of human protein-coding sequences by 30%, from [~]19,500 annotated CDSs to over 26,000. Yet, additional scrutiny of these ORFs has raised numerous questions about what fraction of them truly produce a protein product and what fraction of those can be understood as proteins according to conventional understanding of the term. Adding further complication is the fact that published estimates of non-canonical ORFs vary widely by around 30-fold, from several thousand to several hundred thousand. The summation of this research has left the genomics and proteomics communities both excited by the prospect of new coding regions in the human genome, but searching for guidance on how to proceed. Here, we discuss the current state of non-canonical ORF research, databases, and interpretation, focusing on how to assess whether a given ORF can be said to be "protein-coding". In briefThe human genome encodes thousands of non-canonical open reading frames (ORFs) in addition to protein-coding genes. As a nascent field, many questions remain regarding non-canonical ORFs. How many exist? Do they encode proteins? What level of evidence is needed for their verification? Central to these debates has been the advent of ribosome profiling (Ribo-seq) as a method to discern genome-wide ribosome occupancy, and immunopeptidomics as a method to detect peptides that are processed and presented by MHC molecules and not observed in traditional proteomics experiments. This article provides a synthesis of the current state of non-canonical ORF research and proposes standards for their future investigation and reporting. HighlightsO_LICombined use of Ribo-seq and proteomics-based methods enables optimal confidence in detecting non-canonical ORFs and their protein products. C_LIO_LIRibo-seq can provide more sensitive detection of non-canonical ORFs, but data quality and analytical pipelines will impact results. C_LIO_LINon-canonical ORF catalogs are diverse and span both high-stringency and low-stringency ORF nominations. C_LIO_LIA framework for standardized non-canonical ORF evidence will advance the research field. C_LI Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=107 SRC="FIGDIR/small/541049v1_ufig1.gif" ALT="Figure 1"> View larger version (28K): org.highwire.dtl.DTLVardef@4878d0org.highwire.dtl.DTLVardef@1f3c9e4org.highwire.dtl.DTLVardef@4ba1f7org.highwire.dtl.DTLVardef@17695a6_HPS_FORMAT_FIGEXP M_FIG C_FIG

genomics↗

bioRxiv2022

ARAX: a graph-based modular reasoning tool for translational biomedicine

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSMotivationC_ST_ABSWith the rapidly growing volume of knowledge and data in biomedical databases, improved methods for knowledge-graph-based computational reasoning are needed in order to answer translational questions. Previous efforts to solve such challenging computational reasoning problems have contributed tools and approaches, but progress has been hindered by the lack of an expressive analysis workflow language for translational reasoning and by the lack of a reasoning engine--supporting that language--that federates semantically integrated knowledge-bases. ResultsWe introduce ARAX, a new reasoning system for translational biomedicine that provides a web browser user interface and an application programming interface. ARAX enables users to encode translational biomedical questions and to integrate knowledge across sources to answer the users query and facilitate exploration of results. For ARAX, we developed new approaches to query planning, knowledge-gathering, reasoning, and result ranking and dynamically integrate knowledge providers for answering biomedical questions. To illustrate ARAXs application and utility in specific disease contexts, we present several use-case examples. Availability and ImplementationThe source code and technical documentation for building the ARAX server-side software and its built-in knowledge database are freely available online (https://github.com/RTXteam/RTX). We provide a hosted ARAX service with a web browser interface at arax.rtx.ai and a web application programming interface (API) endpoint at arax.rtx.ai/api/arax/v1.3/ui/. Contactdmk333@psu.edu

bioinformatics↗

bioRxiv2022

A comprehensive evaluation of consensus spectrum generation methods in proteomics

Spectrum clustering is a powerful strategy to minimize redundant mass spectral data by grouping highly similar mass spectra corresponding to repeatedly measured analytes. Based on spectrum similarity, near-identical spectra are grouped in clusters, after which each cluster can be represented by its so-called consensus spectrum for downstream processing. Although several algorithms for spectrum clustering have been adequately benchmarked and tested, the influence of the consensus spectrum generation step is rarely evaluated. Here, we present an implementation and benchmark of common consensus spectrum algorithms, including spectrum averaging, spectrum binning, the most similar spectrum, and the best-identified spectrum. We have analyzed diverse public datasets using two different clustering algorithms (spectra-cluster and MaRaCluster) to evaluate how the consensus spectrum generation procedure influences downstream peptide identification. The BEST and BIN methods were found the most reliable methods for consensus spectrum generation, including for datasets with post-translational modifications (PTM) such as phosphorylation. All source code and data of the present study are freely available on GitHub at https://github.com/statisticalbiotechnology/representative-spectra-benchmark.

bioinformatics↗

bioRxiv2021

A method for independent estimation of false localisation rate for phosphoproteomics

Phosphoproteomics methods are commonly employed in labs to identify and quantify the sites of phosphorylation on proteins. In recent years, various software tools have been developed, incorporating scores or statistics related to whether a given phosphosite has been correctly identified, or to estimate the global false localisation rate (FLR) within a given data set for all sites reported. These scores have generally been calibrated using synthetic data sets, and their statistical reliability on real datasets is largely unknown. As a result, there is considerable problem in the field of reporting incorrectly localised phosphosites, due to inadequate statistical control. In this work, we develop the concept of using scoring and ranking modifications on a decoy amino acid, i.e. one that cannot be modified, to allow for independent estimation of global FLR. We test a variety of different amino acids to act as the decoy, on both synthetic and real data sets, demonstrating that the amino acid selection can make a substantial difference to the estimated global FLR. We conclude that while several different amino acids might be appropriate, the most reliable FLR results were achieved using alanine and leucine as decoys, although we have a preference for alanine due to the risk of potential confusion between leucine and isoleucine amino acids. We propose that the phosphoproteomics field should adopt the use of a decoy amino acid, so that there is better control of false reporting in the literature, and in public databases that re-distribute the data. Data are available via ProteomeXchange with identifier PXD028840.

bioinformatics↗

bioRxiv2021

RTX-KG2: a system for building a semantically standardized knowledge graph for translational biomedicine

BackgroundBiomedical translational science is increasingly using computational reasoning on repositories of structured knowledge (such as UMLS, SemMedDB, ChEMBL, Reactome, DrugBank, and SMPDB in order to facilitate discovery of new therapeutic targets and modalities. The NCATS Biomedical Data Translator project is working to federate autonomous reasoning agents and knowledge providers within a distributed system for answering translational questions. Within that project and the broader field, there is a need for a framework that can efficiently and reproducibly build an integrated, standards-compliant, and comprehensive biomedical knowledge graph that can be downloaded in standard serialized form or queried via a public application programming interface (API). ResultsTo create a knowledge provider system within the Translator project, we have developed RTX-KG2, an open-source software system for building--and hosting a web API for querying--a biomedical knowledge graph that uses an Extract-Transform-Load (ETL) approach to integrate 70 knowledge sources (including the aforementioned core six sources) into a knowledge graph with provenance information including (where available) citations. The semantic layer and schema for RTX-KG2 follow the standard Biolink model to maximize interoperability. RTX-KG2 is currently being used by multiple Translator reasoning agents, both in its downloadable form and via its SmartAPI-registered interface. Serializations of RTX-KG2 are available for download in both the pre-canonicalized form and in canonicalized form (in which synonyms are merged). The current canonicalized version (KG2.7.3) of RTX-KG2 contains 6.4M nodes and 39.3M edges with a hierarchy of 77 relationship types from Biolink. ConclusionRTX-KG2 is the first knowledge graph that integrates UMLS, SemMedDB, ChEMBL, DrugBank, Reactome, SMPDB, and 64 additional knowledge sources within a knowledge graph that conforms to the Biolink standard for its semantic layer and schema. RTX-KG2 is publicly available for querying via its API at arax.rtx.ai/api/rtxkg2/v1.2/openapi.json. The code to build RTX-KG2 is publicly available at github:RTXteam/RTX-KG2.

bioinformatics↗