Biology subjects

Jinich, A.

Publications and source records attributed to Jinich, A..

At least 19 recordsLinked to original sources

bioRxiv2026

ProtoPure: Literature Mining and Analysis of Protein Purification Protocols at Proteome Scale

Protein purification is a fundamental step for analyzing individual proteins and protein complexes and identifying interactions with other proteins, DNA, or RNA. A variety of protein purification strategies exist to address desired scale, throughput, and downstream applications. However, selecting an effective experimental procedure remains a largely empirical process that a quick literature review can help guide. Researchers typically survey on the order of a dozen publications for proteins with similar biochemical properties, assessing conditions on a case-by-case basis, leaving no systematic framework for comparison and decision-making at the scale of entire protein families. As protein engineering and design evolve, identifying effective purification conditions remains critical for obtaining sufficient quantities of properly folded and functional proteins. To address this gap, we developed ProtoPure, an LLM-based workflow and database that automates the extraction, standardization and analysis of protein purification conditions from peer-reviewed literature at scale. Given a user-defined UniProt table-enabling flexible selection of proteins by family, organism, or proteome. ProtoPure retrieves associated publications, classifies them based on the presence of purification protocols, and extracts a standardized set of 12 purification parameters (e.g., buffer composition, elution conditions). By transforming dispersed experimental descriptions into a structured resource, ProtoPure makes it possible to compare purification strategies across thousands of proteins. The resulting dataset revealed quantitative convergence toward a limited set of recombinant expression practices, with BL21(DE3)-based systems dominating protein production and recurrent combinations of expression host, inducer, growth medium, and plasmid occurring substantially more frequently than expected by chance. In contrast, downstream purification parameters exhibited greater diversity, revealing a workflow architecture that combines standardized expression infrastructure with more flexible biochemical components.

biophysics↗

bioRxiv2026

De novo minibinders targeting RH5 achieve nanomolar inhibition of blood-stage malaria replication

Malaria kills {approx}600,000 people annually despite recently deployed vaccines. The parasite's blood-stage invasion of erythrocytes, driven by the essential, non-redundant interaction between Plasmodium falciparum RH5 and erythrocyte basigin, remains a leading therapeutic target. Here, we designed de novo mini-proteins (minibinders) to occlude this interface. From {approx}1,500 computationally generated designs, we synthesized 18 candidates ranging from 9.3 to 16.9 kDa which were easily expressed in E. coli as soluble, well-folded proteins. Of these, 7 minibinders blocked blood-stage replication of P. falciparum asexual stage parasites with greater potency than {approx}300nM IC50. Three potent candidate minibinders, mb-5, mb-7, and mb-21, were confirmed to bind RH5 via SPR and BLI and block the merozoite-to-ring transition in stage-synchronized parasite progression assays, with CryoEM of minibinder-RH5 complex confirming the predicted binding at RH5-basigin interface. Although pharmacological challenges remain, our data demonstrate that potent inhibitors of parasite growth, rivaling existing antimalarials, can be designed rather than discovered via extensive screening. Our data highlight the potential for artificial intelligence to change the development landscape of new antimalarial therapies.

microbiology↗

bioRxiv2026

Alkamines reveal a hidden layer of steroid and drug metabolism

Biomedical research overlooks most genes in favor of a well-studied minority, yet whether analogous blind spots exist in metabolomics remains unknown. We show that reductive amination, forming secondary amines from aldehydes or ketones and amines, generates a previously hidden class of metabolites we term alkamines. Multiplexed synthesis of 8,475 alkamines combined with MS/MS searches across 1.7 billion spectra identified 1,626 candidates across multiple species and organs. Of these, 56 were confirmed in biological samples, including 27 steroid- and 12 drug-derived alkamines matching prescription patterns. Notably, 77% of synthesized alkamines are absent from PubChem. This combinatorial logic likely explains why alkamines have evaded detection and suggests drug metabolism frameworks substantially underestimate drug-derived metabolite diversity. Reductive amination is an overlooked route modifying steroids, bile acids, and xenobiotics.

biochemistry↗

bioRxiv2026

TubercuProbe: A Cross-Attention Graph-Sequence Model for Cross-Species Chemoproteomic Discovery in Mycobacterium tuberculosis

Activity-based protein profiling (ABPP) and residue-specific chemoproteomics have transformed human chemical biology, yet applying these approaches to pathogens remains limited by biosafety constraints, low throughput, and the absence of reusable atlases. We present TubercuProbe, a cross-species machine learning framework that leverages large-scale human chemoproteomic knowledge to prioritize compound-protein interactions in Mycobacterium tuberculosis (Mtb) and other pathogens. Our model integrates a graph isomorphism network (GINE) for ligand encoding with frozen ESM-C (600M) protein embeddings via bidirectional cross-attention. Trained on >2M ChEMBL compound-protein pairs (predominantly human targets), TubercuProbe achieves R2=0.77 (MSE=0.45) for continuous affinity prediction and transfers effectively to binary cysteine reactivity prediction (CysDB AUPRC=0.63). Ablation studies reveal that pretrained features are highly transferable across freezing strategies ({Delta}AUPRC<0.03), suggesting the model captures fundamental protein-molecule interaction patterns. As a case study, we prioritize cysteine-reactive electrophiles and molecular glues for three Mtb virulence proteins (PtpB, SapM, Rv3671c), providing candidate probes for prospective ABPP validation. Orthogonal comparison with Boltz-2 structure predictions shows moderate correlation (Pearson r{approx}0.69). TubercuProbe provides a lightweight, sequence-driven first-pass ranker that enables pre-experimental prioritization--reducing the time, cost, and experimental burden of chemoproteomic discovery in biosafety-restricted systems. We discuss extensions toward multitask learning that jointly predicts non-covalent binding and covalent reactivity, recognizing that effective covalent probes must both reach their target and react once there.

biochemistry↗

bioRxiv2025

Inferring Local Protein Structural Similarity from Sequence Alone

Detecting structural similarity at the local level between proteins is central to understanding function and evolution, yet most approaches require 3D models. In this work, we show that protein language models (pLMs), solely using sequence data as input, implicitly capture fine-grained structural signals that can be leveraged to identify such similarities. By mean-pooling residue embeddings over sliding windows and comparing them across proteins with cosine similarity, we find diagonal patterns that reflect locally aligned regions even without sequence identity. Building on this insight, we introduce a framework for detecting locally aligned structural regions directly from sequences, supporting the development of scalable methods for structural annotation and comparison.

bioinformatics↗

bioRxiv2025

Genes required for Mycobacterium tuberculosis to survive the transition from aerosol to pulmonary alveolar lining fluid and early infection in a model of transmission

Mycobacterium tuberculosis (Mtb) must survive multiple changes in environment for aerosol transmission. Our genome-wide screen for rehydration of modeled aerosols in surrogate pulmonary alveolar lining fluid identified a survival-sustaining role for 22 genes not required for survival during earlier stages of transmission, 20 of which are non-essential in routine culture. Thirteen genes were also needed for Mtb to survive in alveolar macrophages after passing through the earlier stages of transmission. Nine 9 of the 13 were needed for full infectivity of aerosols in mice. Seven of the genes sustaining the intra-alveolar survival of aerosolized Mtb are likely to regulate Mtbs uptake, catabolism or synthesis of lipids as regulated by cAMP. Our results reveal a dynamic form of conditional essentiality that only emerges after bacteria experience a sequence of other pathophysiologically relevant conditions. These findings enlarge Mtbs candidate transmission survival genome with stage-specific genes encoding potential targets for blocking tuberculosis transmission.

microbiology↗

bioRxiv2025

A periplasmic protein complex mediates cell envelope integrity and intrinsic multidrug resistance in Mycobacterium tuberculosis

The intrinsic drug resistance of Mycobacterium tuberculosis (Mtb) is a major barrier to effective tuberculosis (TB) treatment, largely due to its complex, impermeable cell envelope. We identified a periplasmic protein complex comprising FecB and Rv3035 that is essential for maintaining envelope integrity and mediating intrinsic multidrug resistance in Mtb. FecB interacts with Rv3035, forming a stable heterodimer that associates with the cell envelope biosynthesis protein AftB. We report the structures of Rv3035 alone and in complex with FecB and identify critical residues for complex formation and function. Co-essentiality and genetic interaction analyses support a functional link between FecB, Rv3035 and AftB, an arabinofuranosyltransferase which synthesizes arabinogalactan and lipoarabinomannan. Loss of FecB or Rv3035 disrupted AftB-mediated arabinan synthesis, suggesting that these proteins support AftBs enzymatic activity. Importantly, FecB is required for Mtb virulence in mice, underscoring its physiological relevance. These findings highlight FecB, Rv3035 and AftB as promising therapeutic targets.

microbiology↗

bioRxiv2025

Candidate transmission survival genome of Mycobacterium tuberculosis

Mycobacterium tuberculosis (Mtb), a leading cause of death from infection, completes its life cycle entirely in humans except for transmission through the air. To begin to understand how Mtb survives aerosolization, we mimicked liquid and atmospheric conditions experienced by Mtb before and after exhalation using a model aerosol fluid (MAF) based on the water-soluble, lipidic and cellular constituents of necrotic tuberculosis lesions. MAF induced drug tolerance in Mtb, remodeled its transcriptome and protected Mtb from dying in microdroplets desiccating in air. Yet survival was not passive: Mtb appeared to rely on hundreds of genes to survive conditions associated with transmission. Essential genes subserving proteostasis offered most protection. A large number of conventionally nonessential genes appeared to contribute as well, including genes encoding proteins that resemble anti-desiccants. The candidate transmission survival genome of Mtb may offer opportunities to reduce transmission of tuberculosis. Significance StatementMycobacterium tuberculosis (Mtb) travels from the lungs of one person through the air to the lungs of another and survives multiple stresses en route, including changes in temperature and in concentrations of oxygen, carbon dioxide, hydrogen ions, salts and organic solutes. Here we present a genetically tractable model of transmission to begin the identification of the transmission survival genome of Mtb. We devised a fluid that mimics TB lesions, found that it protects Mtb from transmission-related stresses, associated this with the structure of the droplets as they dry and their ability to retain water, and used it to query the potential contribution of each of Mtbs genes to Mtbs survival in models of three sequential stages of transmission.

microbiology↗

bioRxiv2024

The Protein Language Visualizer: Sequence Similarity Networks for the Era of Language Models

AO_SCPLOWBSTRACTC_SCPLOWThe era of modern AI-driven representations of proteins is here, and moving fast, yet tools for their intuitive visualization and exploration lag behind. Sequence Similarity Networks (SSNs) have long filled this role for alignment-based methods, providing simple but widely adopted platforms for grouping proteins by homology. Building on this foundation, we present the Protein Language Visualizer (PLVis), a modular framework that applies existing pre-trained protein language model (pLM) embeddings, dimensionality reduction, and clustering to generate interactive maps of protein relationships. The central contribution is the PLVis Repository, an online resource where thousands of reference proteomes can be compared and annotated through an accessible, interactive interface, much like SSNs became impactful not for their technical novelty but for their broad usability. We first validate that well-separated clusters in PLVis reliably capture homology information, while emphasizing caution when interpreting central "fuzzy" regions. We then illustrate the value of PLVis through case studies spanning individual protein families to full proteome comparisons across Mycobacterium and Plasmodium species. By combining methodological clarity with broad accessibility, the PLVis Repository provides a low-barrier platform for exploring proteomes through the lens of language models.

bioinformatics↗

bioRxiv2024

Annotation-free prediction of microbial dioxygen utilization

Aerobes require dioxygen (O2) to grow; anaerobes do not. But nearly all microbes -- aerobes, anaerobes, and facultative organisms alike -- express enzymes whose substrates include O2, if only for detoxification. This presents a challenge when trying to assess which organisms are aerobic from genomic data alone. This challenge can be overcome by noting that O2 utilization has wide-ranging effects on microbes: aerobes typically have larger genomes, encode more O2-utilizing enzymes, and tend to use different amino acids in their proteins. Here we show that these effects permit high-quality prediction of O2 utilization from genome sequences, with several models displaying >70% balanced accuracy on a ternary classification task wherein blind guessing is only 33.3% accurate. Since genome annotation is compute-intensive and relies on many assumptions, we asked if annotation-free methods also perform well. We discovered that simple and efficient models based entirely on genome sequence content -- e.g. triplets of amino acids -- perform about as well as intensive annotation-based algorithms, enabling the rapid processing of global-scale sequence data to predict aerobic physiology. To demonstrate the utility of efficient physiological predictions we estimated the prevalence of aerobes and anaerobes along a well-studied O2 gradient in the Black Sea, finding strong quantitative correspondence between local chemistry (O2:sulfide concentration ratio) and the composition of microbial communities. We therefore suggest that statistical methods like ours can be used to estimate, or "sense," pivotal features of the environment from DNA sequencing data. ImportanceWe now have access to sequence data from a wide variety of natural environments. These data document a bewildering diversity of microbes, many known only from their genomes. Physiology -- an organisms capacity to engage metabolically with its environment -- may provide a more useful lens than taxonomy for understanding microbial communities. As an example of this broader principle, we developed algorithms that accurately predict microbial dioxygen utilization directly from genome sequences without first annotating genes, e.g. by considering only the amino acids in protein sequences. Annotation-free algorithms enabled rapid characterization of natural samples, demonstrating a quantitative correspondence between sequences and local O2 levels. These results suggest that DNA sequencing can be repurposed as a multi-pronged chemical sensor, estimating concentrations of O2 and other key facets of complex natural settings.

microbiology↗

bioRxiv2023

Genome-wide screen identifies host loci that modulate M. tuberculosis fitness in immunodivergent mice

Genetic differences among mammalian hosts and Mycobacterium tuberculosis (Mtb) strains determine diverse tuberculosis (TB) patient outcomes. The advent of recombinant inbred mouse panels and next-generation transposon mutagenesis and sequencing approaches has enabled dissection of complex host- pathogen interactions. To identify host and pathogen genetic determinants of Mtb pathogenesis, we infected members of the BXD family of mouse strains with a comprehensive library of Mtb transposon mutants (TnSeq). Members of the BXD family segregate for Mtb-resistant C57BL/6J (B6 or B) and Mtb-susceptible DBA/2J (D2 or D) haplotypes. The survival of each bacterial mutant was quantified within each BXD host, and we identified those bacterial genes that were differentially required for Mtb fitness across BXD genotypes. Mutants that varied in survival among the host family of strains were leveraged as reporters for "endophenotypes", each bacterial fitness profile directly probing specific components of the infection microenvironment. We conducted QTL mapping of these bacterial fitness endophenotypes and identified 140 host-pathogen quantitative trait loci (hpQTL). We identified a QTL hotspot on chromosome 6 (75.97-88.58 Mb) associated with the genetic requirement of multiple Mtb genes; Rv0127 (mak), Rv0359 (rip2), Rv0955 (perM), and Rv3849 (espR). Together, this screen reinforces the utility of bacterial mutant libraries as precise reporters of the host immunological microenvironment during infection and highlights specific host-pathogen genetic interactions for further investigation. To enable downstream follow-up for both bacterial and mammalian genetic research communities, all bacterial fitness profiles have been deposited into GeneNetwork.org and added into the comprehensive collection of TnSeq libraries in MtbTnDB.

genetics↗

bioRxiv2022

Predicting enzyme substrate chemical structure with protein language models

The number of unannotated or orphan enzymes vastly outnumber those for which the chemical structure of the substrates are known. While a number of enzyme function prediction algorithms exist, these often predict Enzyme Commission (EC) numbers or enzyme family, which limits their ability to generate experimentally testable hypotheses. Here, we harness protein language models, cheminformatics, and machine learning classification techniques to accelerate the annotation of orphan enzymes by predicting their substrates chemical structural class. We use the orphan enzymes of Mycobacterium tuberculosis as a case study, focusing on two protein families that are highly abundant in its proteome: the short-chain dehydrogenase/reductases (SDRs) and the S-adenosylmethionine (SAM)-dependent methyltransferases. Training machine learning classification models that take as input the protein sequence embeddings obtained from a pre-trained, self-supervised protein language model results in excellent accuracy for a wide variety of prediction tasks. These include redox cofactor preference for SDRs; small-molecule vs. polymer (i.e. protein, DNA or RNA) substrate preference for SAM-dependent methyltransferases; as well as more detailed chemical structural predictions for the preferred substrates of both enzyme families. We then use these trained classifiers to generate predictions for the full set of unannotated SDRs and SAM-methyltransferases in the proteomes of M. tuberculosis and other mycobacteria, generating a set of biochemically testable hypotheses. Our approach can be extended and generalized to other enzyme families and organisms, and we envision it will help accelerate the annotation of a large number of orphan enzymes. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=106 SRC="FIGDIR/small/509940v3_ufig1.gif" ALT="Figure 1"> View larger version (20K): org.highwire.dtl.DTLVardef@1dab89forg.highwire.dtl.DTLVardef@8eec65org.highwire.dtl.DTLVardef@141ff13org.highwire.dtl.DTLVardef@1d16212_HPS_FORMAT_FIGEXP M_FIG C_FIG

microbiology↗

bioRxiv2022

Genome-wide co-essentiality analysis in Mycobacterium tuberculosis reveals an itaconate defense enzyme module

Genome-wide random mutagenesis screens using transposon sequencing (TnSeq) have been a cornerstone of functional genetics in Mycobacterium tuberculosis (Mtb), helping to define gene essentiality across a wide range of experimental conditions. Here, we harness a recently compiled TnSeq database to identify pairwise correlations of gene essentiality profiles (i.e. co-essentiality analysis) across the Mtb genome and reveal clusters of genes with similar function. We describe selected modules identified by our pipeline, review the literature supporting their associations, and propose hypotheses about novel associations. We focus on a cluster of seven enzymes for experimental validation, characterizing it as an enzymatic arsenal that helps Mtb counter the toxic effects of itaconate, a host-derived antibacterial compound. We extend the use of these correlations to enable prediction of protein complexes by designing a virtual screen that ranks potentially interacting heterodimers from co-essential protein pairs. We envision co-essentiality analysis will help accelerate gene functional discovery in this important human pathogen.

microbiology↗

bioRxiv2022

Exploring protein sequence similarity with Protein Language UMAPs (PLUMAPs)

Visualizing relationships and similarities between proteins can reveal insightful biology. Current approaches to visualize and analyze proteins based on sequence homology, such as sequence similarity networks (SSNs), create representations of BLAST-based pairwise comparisons. These approaches could benefit from incorporating recent protein language models, which generate high-dimensional vector representations of protein sequences from self-supervised learning on hundreds of millions of proteins. Inspired by SSNs, we developed an interactive tool - Protein Language UMAPs (PLUMAPs) - to visualize protein similarity with protein language models, dimensionality reduction, and topic modeling. As a case study, we compare our tool to Sequence Similarity Network (SSN) using the proteomes of two related bacterial species, Mycobacterium tuberculosis and Mycobacterium smegmatis. Both SSNs and PLUMAPs generate protein clusters corresponding to protein families and highlight enrichment or depletion across species. However, only in PLUMAPs does the layout distance between proteins and protein clusters meaningfully reflect similarity. Thus in PLUMAPs, related protein families are displayed as nearby clusters, and larger-scale structures correlate with cellular localization. Finally, we adapt techniques from topic modeling to automatically annotate protein clusters, making them more easily interpretable and potentially insightful. We envision that as large protein language models permeate bioinformatics and interactive sequence analysis tools, PLUMAPs will become a useful visualization resource across a wide variety of biological disciplines. Anticipating this, we provide a prototype for an online, open source version of PLUMAPs.

bioinformatics↗

bioRxiv2022

Protein Function Prediction from Three-Dimensional Feature Representations Using Space-Filling Curves

Compact and interpretable structural feature representations are required for accurately predicting properties and the function of proteins. In this work, we construct and evaluate three-dimensional feature representations of protein structures based on space-filling curves. We focus on the problem of enzyme substrate prediction, using two ubiquitous enzyme families as case studies: the short-chain dehydrogenase/reductases (SDRs) and the S-adenosylmethionine dependent methyltransferases (SAM-MTases). Space-filling curves such as Hilbert curve and the Morton curve generate a reversible mapping from discretized three-dimensional to one-dimensional representations and thus help to encode three-dimensional molecular structures in a system-independent way and with a minimal number of parameters. Using three-dimensional structures of SDRs and SAM-MTases generated using AlphaFold2, we assess the performance of the SFC-based feature representations in predictions on a new benchmark database of enzyme classification tasks including their cofactor and substrate selectivity. Gradient-boosted tree classifiers yield binary prediction accuracy of 0.766-0.906 and AUC (area under curve) parameters of 0.828-0.922 for the classification tasks. We investigate the effects of amino acid encoding, spatial orientation, and (the few) parameters of SFC-based encodings on the accuracy of the predictions. Our results suggest that geometry-based approaches such as SFCs are promising for generating protein structural representations and are complementary to the highly parametric methods, for example, convolutional neural networks (CNNs).

bioinformatics↗

bioRxiv2021

Multiple acyl-CoA dehydrogenase deficiency kills Mycobacterium tuberculosis in vitro and during infection

The human pathogen Mycobacterium tuberculosis (Mtb) devotes a significant fraction of its genome to fatty acid metabolism. Although Mtb depends on host fatty acids as a carbon source, fatty acid {beta}-oxidation is mediated by genetically redundant enzymes, which has hampered the development of antitubercular drugs targeting this metabolic pathway. Here, we identify rv0338c, referred to as etfDMtb, to encode a membrane dehydrogenase essential for fatty acid {beta}-oxidation in Mtb. An etfD deletion mutant ({Delta}etfD) was incapable of growing on fatty acids in vitro, with long-chain fatty acids being bactericidal, and failed to grow and survive in mice. The {Delta}etfD metabolome revealed a block in {beta}-oxidation at the step catalyzed by acyl-CoA dehydrogenases (ACADs). In many organisms, including humans, ACADs are functionally dependent on an electron transfer flavoprotein (ETF) and cognate dehydrogenase. Immunoprecipitation identified EtfD in complex with FixA (EtfBMtb). FixA (EtfBMtb) and FixB (EtfAMtb) are homologous to the human ETF subunits. Our results demonstrate that EtfBAMtb constitutes Mtbs ETF, while EtfDMtb, although not homologous to human EtfD, functions as the dehydrogenase. These findings identify Mtbs fatty acid {beta}-oxidation as a novel potential target for TB drug development.

microbiology↗

bioRxiv2021

The Mycobacterium tuberculosis transposon sequencing database (MtbTnDB): a large-scale guide to genetic conditional essentiality

Characterizing genetic essentiality across various conditions is fundamental for understanding gene function. Transposon sequencing (TnSeq) is a powerful technique to generate genome-wide essentiality profiles in bacteria and has been extensively applied to Mycobacterium tuberculosis (Mtb). Dozens of TnSeq screens have yielded valuable insights into the biology of Mtb in vitro, inside macrophages, and in model host organisms. Despite their value, these Mtb TnSeq profiles have not been standardized or collated into a single, easily searchable database. This results in significant challenges when attempting to query and compare these resources, limiting our ability to obtain a comprehensive and consistent understanding of genetic conditional essentiality in Mtb. We address this problem by building a central repository of publicly available Mtb TnSeq screens, the Mtb transposon sequencing database (MtbTnDB). The MtbTnDB is a living resource that encompasses to date {approx}150 standardized TnSeq screens, enabling open access to data, visualizations, and functional predictions through an interactive web app (www.mtbtndb.app). We conduct several statistical analyses on the complete database, such as demonstrating that (i) genes in the same genomic neighborhood have similar TnSeq profiles, and (ii) clusters of genes with similar TnSeq profiles are enriched for genes from similar functional categories. We further analyze the performance of machine learning models trained on TnSeq profiles to predict functional annotation of orphan genes in Mtb. By facilitating the comparison of TnSeq screens across conditions, the MtbTnDB will accelerate the exploration of conditional genetic essentiality, provide insights into the functional organization of Mtb genes, and help predict gene function in this important human pathogen.

microbiology↗

bioRxiv2018

The thermodynamic landscape of carbon redox biochemistry

Redox biochemistry plays a key role in the transduction of chemical energy in living systems. However, the compounds observed in metabolic redox reactions are a minuscule fraction of chemical space. It is not clear whether compounds that ended up being selected as metabolites display specific properties that distinguish them from non-biological compounds. Here we introduce a systematic approach for comparing the chemical space of all possible redox states of linear-chain carbon molecules to the corresponding metabolites that appear in biology. Using cheminformatics and quantum chemistry, we analyze the physicochemical and thermodynamic properties of the biological and non-biological compounds. We find that, among all compounds, aldose sugars have the highest possible number of redox connections to other molecules. Metabolites are enriched in carboxylic acid functional groups and depleted of carbonyls, and have higher solubility than non-biological compounds. Upon constructing the energy landscape for the full chemical space as a function of pH and electron donor potential, we find that over a large range of conditions metabolites tend to have lower Gibbs energies than non-biological molecules. Finally, we generate Pourbaix phase diagrams that serve as a thermodynamic atlas to indicate which compounds are local and global energy minima in redox chemical space across a set of pH values and electron donor potentials. Our work yields insight into the physicochemical principles governing redox metabolism, and suggests that thermodynamic stability in aqueous environments may have played an important role in early metabolic processes.

systems biology↗