Systematic comparison of ranking aggregation methods for gene lists in experimental results.

MOTIVATION: A common experimental output in biomedical science is a list of genes implicated in a given biological process or disease. The gene lists resulting from a group of studies answering the same, or similar, questions can be combined by ranking aggregation methods to find a consensus or a more reliable answer. Evaluating a ranking aggregation method on a specific type of data before using it is required to support the reliability since the property of a dataset can influence the performance of an algorithm. Such evaluation on gene lists is usually based on a simulated database because of the lack of a known truth for real data. However, simulated datasets tend to be too small compared to experimental data and neglect key features, including heterogeneity of quality, relevance and the inclusion of unranked lists. RESULTS: In this study, a group of existing methods and their variations that are suitable for meta-analysis of gene lists are compared using simulated and real data. Simulated data were used to explore the performance of the aggregation methods as a function of emulating the common scenarios of real genomic data, with various heterogeneity of quality, noise level and a mix of unranked and ranked data using 20 000 possible entities. In addition to the evaluation with simulated data, a comparison using real genomic data on the SARS-CoV-2 virus, cancer (non-small cell lung cancer) and bacteria (macrophage apoptosis) was performed. We summarize the results of our evaluation in a simple flowchart to select a ranking aggregation method, and in an automated implementation using the meta-analysis by information content algorithm to infer heterogeneity of data quality across input datasets. AVAILABILITY AND IMPLEMENTATION: The code for simulated data generation and running edited version of algorithms: https://github.com/baillielab/comparison_of_RA_methods. Code to perform an optimal selection of methods based on the results of this review, using the MAIC algorithm to infer the characteristics of an input dataset, can be downloaded here: https://github.com/baillielab/maic. An online service for running MAIC: https://baillielab.net/maic. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Molecular Patterns in Acute Pancreatitis Reflect Generalizable Endotypes of the Host Response to Systemic Injury in Humans.

OBJECTIVE: Acute Pancreatitis (AP) is sudden onset pancreas inflammation that causes systemic injury with a wide and markedly heterogeneous range of clinical consequences. Here, we hypothesized that this observed clinical diversity corresponds to diversity in molecular subtypes that can be identified in clinical and multiomics data. SUMMARY BACKGROUND DATA: Observational cohort study. n = 57 for the discovery cohort (clinical, transcriptomics, proteomics, and metabolomics data) and n = 312 for the validation cohort (clinical and metabolomics data). METHODS: We integrated coincident transcriptomics, proteomics, and metabolomics data at serial time points between admission to hospital and up to 48 hours after recruitment from a cohort of patients presenting with acute pancreatitis. We systematically evaluated 4 different metrics for patient similarity using unbiased mathematical, biological, and clinical measures of internal and external validity.We next compared the AP molecular endotypes with previous descriptions of endotypes in a critically ill population with acute respiratory distress syndrome (ARDS). RESULTS: Our results identify 4 distinct and stable AP molecular endotypes. We validated our findings in a second independent cohort of patients with AP.We observed that 2 endotypes in AP recapitulate disease endotypes previously reported in ARDS. CONCLUSIONS: Our results show that molecular endotypes exist in AP and reflect biological patterns that are also present in ARDS, suggesting that generalizable patterns exist in diverse presentations of critical illness.

Cisplatin Uptake in Macrophage Subtypes at the Single-Cell Level by LA-ICP-TOFMS Imaging.

A high-throughput laser ablation-inductively coupled plasma-time-of-flight mass spectrometry (LA-ICP-TOFMS) workflow was implemented for quantitative single-cell analysis following cytospin preparation of cells. For the first time, in vitro studies on cisplatin exposure addressed human monocytes and monocyte-derived macrophages (undifferentiated THP-1 monocytic cells, differentiated M0 macrophages, as well as further polarized M1 and M2 phenotypes) at the single-cell level. The models are of particular interest as macrophages comprise the biggest part of immune cells present in the tumor microenvironment and play an important role in modulating tumor growth and progression. The introduced bioimaging workflow proved to be universally applicable to adherent and suspension cell cultures and fit-for-purpose for the quantitative analysis of several hundreds of cells within minutes. Both, cross-validation of the method with single-cell analysis in suspension for THP-1 cells and with LA-ICP-TOFMS analysis of adherent M0 cells grown on chambered glass coverslips, revealed agreeing platinum concentrations at the single-cell level. A high incorporation of cisplatin was observed in M2 macrophages compared to the M0 and M1 macrophage subtypes and the monocyte model, THP-1. The combination with bright-field images and monitoring of highly abundant endogenous elements such as phosphorus and sodium at a high spatial resolution allowed assessing cell size and important morphological cell parameters and thus straightforward control over several cell conditions. This way, apoptotic cells and cell debris as well as doublets or cell clusters could be easily excluded prior to data evaluation without additional staining.

Fundamentals and Recent Developments in Approximate Bayesian Computation.

Bayesian inference plays an important role in phylogenetics, evolutionary biology, and in many other branches of science. It provides a principled framework for dealing with uncertainty and quantifying how it changes in the light of new evidence. For many complex models and inference problems, however, only approximate quantitative answers are obtainable. Approximate Bayesian computation (ABC) refers to a family of algorithms for approximate inference that makes a minimal set of assumptions by only requiring that sampling from a model is possible. We explain here the fundamentals of ABC, review the classical algorithms, and highlight recent developments. [ABC; approximate Bayesian computation; Bayesian inference; likelihood-free inference; phylogenetics; simulator-based models; stochastic simulation models; tree-based models.]

Simultaneous Estimation of Nongaussian Components and Their Correlation Structure.

The statistical dependencies that independent component analysis (ICA) cannot remove often provide rich information beyond the linear independent components. It would thus be very useful to estimate the dependency structure from data. While such models have been proposed, they have usually concentrated on higher-order correlations such as energy (square) correlations. Yet linear correlations are a fundamental and informative form of dependency in many real data sets. Linear correlations are usually completely removed by ICA and related methods so they can only be analyzed by developing new methods that explicitly allow for linearly correlated components. In this article, we propose a probabilistic model of linear nongaussian components that are allowed to have both linear and energy correlations. The precision matrix of the linear components is assumed to be randomly generated by a higher-order process and explicitly parameterized by a parameter matrix. The estimation of the parameter matrix is shown to be particularly simple because using score-matching (Hyvärinen, 2005 ), the objective function is a quadratic form. Using simulations with artificial data, we demonstrate that the proposed method improves the identifiability of nongaussian components by simultaneously learning their correlation structure. Applications on simulated complex cells with natural image input, as well as spectrograms of natural audio data, show that the method finds new kinds of dependencies between the components.

The impact of host metapopulation structure on the population genetics of colonizing bacteria.

Many key bacterial pathogens are frequently carried asymptomatically, and the emergence and spread of these opportunistic pathogens can be driven, or mitigated, via demographic changes within the host population. These inter-host transmission dynamics combine with basic evolutionary parameters such as rates of mutation and recombination, population size and selection, to shape the genetic diversity within bacterial populations. Whilst many studies have focused on how molecular processes underpin bacterial population structure, the impact of host migration and the connectivity of the local populations has received far less attention. A stochastic neutral model incorporating heightened local transmission has been previously shown to fit closely with genetic data for several bacterial species. However, this model did not incorporate transmission limiting population stratification, nor the possibility of migration of strains between subpopulations, which we address here by presenting an extended model. We study the consequences of migration in terms of shared genetic variation and show by simulation that the previously used summary statistic, the allelic mismatch distribution, can be insensitive to even large changes in microepidemic and migration rates. Using likelihood-free inference with genotype network topological summaries we fit a simpler model to commensal and hospital samples from the common nosocomial pathogens Staphylococcus aureus, Staphylococcus epidermidis, Enterococcus faecalis and Enterococcus faecium. Only the hospital data for E. faecium display clearly marked deviations from the model predictions which may be attributable to its adaptation to the hospital environment.

Direct learning of sparse changes in Markov networks by density ratio estimation.

We propose a new method for detecting changes in Markov network structure between two sets of samples. Instead of naively fitting two Markov network models separately to the two data sets and figuring out their difference, we directly learn the network structure change by estimating the ratio of Markov network models. This density-ratio formulation naturally allows us to introduce sparsity in the network structure change, which highly contributes to enhancing interpretability. Furthermore, computation of the normalization term, a critical bottleneck of the naive approach, can be remarkably mitigated. We also give the dual formulation of the optimization problem, which further reduces the computation cost for large-scale Markov networks. Through experiments, we demonstrate the usefulness of our method.

Designing optimal behavioral experiments using machine learning.

Computational models are powerful tools for understanding human cognition and behavior. They let us express our theories clearly and precisely and offer predictions that can be subtle and often counter-intuitive. However, this same richness and ability to surprise means our scientific intuitions and traditional tools are ill-suited to designing experiments to test and compare these models. To avoid these pitfalls and realize the full potential of computational modeling, we require tools to design experiments that provide clear answers about what models explain human behavior and the auxiliary assumptions those models must make. Bayesian optimal experimental design (BOED) formalizes the search for optimal experimental designs by identifying experiments that are expected to yield informative data. In this work, we provide a tutorial on leveraging recent advances in BOED and machine learning to find optimal experiments for any kind of model that we can simulate data from, and show how by-products of this procedure allow for quick and straightforward evaluation of models and their parameters against real experimental data. As a case study, we consider theories of how people balance exploration and exploitation in multi-armed bandit decision-making tasks. We validate the presented approach using simulations and a real-world experiment. As compared to experimental designs commonly used in the literature, we show that our optimal designs more efficiently determine which of a set of models best account for individual human behavior, and more efficiently characterize behavior given a preferred model. At the same time, formalizing a scientific question such that it can be adequately addressed with BOED can be challenging and we discuss several potential caveats and pitfalls that practitioners should be aware of. We provide code to replicate all analyses as well as tutorial notebooks and pointers to adapt the methodology to different experimental settings.

Clofarabine Has a Superior Therapeutic Window as compared to Gemcitabine in Preclinical Bladder Cancer Models.

Current standard-of-care systemic therapy options for locally advanced and metastatic bladder cancer (BC), which are predominantly based on cisplatin-gemcitabine combinations, are limited by significant treatment failure rates and frailty-based patient ineligibility. We previously addressed the urgent clinical need for better-tolerated BC therapeutic strategies using a drug screening approach, which identified outstanding antineoplastic activity of clofarabine in preclinical models of BC. To further assess clofarabine as a potential BC therapy component, we conducted head-to-head comparisons of responses to clofarabine versus gemcitabine in preclinical in vitro and in vivo models of BC, complemented by in silico analyses. In vitro data suggest a distinct correlation between the two antimetabolites, with higher cytotoxicity of gemcitabine, especially against several nonmalignant cell types, including keratinocytes and endothelial cells. Accordingly, tolerance of clofarabine (oral or intraperitoneal application) was distinctly better than for gemcitabine (intraperitoneal) in patient-derived xenograft models of BC. Clofarabine also exhibited distinctly superior anticancer efficacy, even at dosing regimens optimized for gemcitabine. Neither complete remission nor cure, both of which were observed with clofarabine, were achieved with any tolerable gemcitabine regimen. Taken together, our findings demonstrate that clofarabine has a better therapeutic window than gemcitabine, further emphasizing its potential as a candidate for drug repurposing in BC. PATIENT SUMMARY: We compared the anticancer activity of clofarabine, a drug used for treatment of leukemia but not bladder cancer, and gemcitabine, a drug currently used for chemotherapy against bladder cancer. Using cell cultures and mouse models, we found that clofarabine was better tolerated and more efficacious than gemcitabine, and even cured implanted tumors in mouse models. Our results suggest that clofarabine, alone or in combination schemes, might be superior to gemcitabine for the treatment of bladder cancer.

Differentiated and non-differentiated HepaRG™ cells: A possible in-vitro model system for early hepatocarcinogenesis and non-genotoxic carcinogens.

Many xenobiotics are non-genotoxic carcinogens (NGC) in rodent liver. Their mode of action (MoA) and health risks for humans are unclear and no in-vitro tests are available to predict NGC. Human HepaRG™ cells in the differentiated (d-HepaRG) and non-differentiated state (nd-HepaRG) were studied as new approach methodology (NAM) for NGC. Cell-biological assays were performed with d-/nd-HepaRG and human hepatoma/hepatocarcinoma cell lines to characterize the benign/malignant phenotype. Reaction of d-/nd-HepaRG to several liver growth factors and NGC (phenobarbital, PB; cyproterone acetate, CPA; WY-14643) was compared to unaltered and premalignant rat hepatocytes in ex-vivo culture. Enzyme induction by NGC was checked by RT-qPCR/oligo-arrays. Growth, anchorage-independency, migration, clonogenicity, and in-vivo tumorigenicity of nd-HepaRG ranged between benign d-HepaRG and malignant hepatoma/hepatocarcinoma cells. All growth factors elevated DNA replication of d-/nd-HepaRG cells, similarly to unaltered/premalignant rat hepatocytes. NGC induced their prototypical enzymes in the rat and human cells, but elicited a growth response only in the unaltered/premalignant rat hepatocytes and not in human d-/nd-HepaRG cells. To conclude, a benign/premalignant phenotype of d-/nd-HepaRG cells and a reactivity towards several hepatic growth factors and NGC, as known from human hepatocytes, are essential components for an in-vitro model for early stage human hepatocarcinogenesis.The potential value as new approach methodology (NAM) for NGC is discussed.

Inference and uncertainty quantification of stochastic gene expression via synthetic models.

Estimating uncertainty in model predictions is a central task in quantitative biology. Biological models at the single-cell level are intrinsically stochastic and nonlinear, creating formidable challenges for their statistical estimation which inevitably has to rely on approximations that trade accuracy for tractability. Despite intensive interest, a sweet spot in this trade-off has not been found yet. We propose a flexible procedure for uncertainty quantification in a wide class of reaction networks describing stochastic gene expression including those with feedback. The method is based on creating a tractable coarse-graining of the model that is learned from simulations, a synthetic model, to approximate the likelihood function. We demonstrate that synthetic models can substantially outperform state-of-the-art approaches on a number of non-trivial systems and datasets, yielding an accurate and computationally viable solution to uncertainty quantification in stochastic models of gene expression.

The copper transporter CTR1 and cisplatin accumulation at the single-cell level by LA-ICP-TOFMS.

More than a decade ago, studies on cellular cisplatin accumulation via active membrane transport established the role of the high affinity copper uptake protein 1 (CTR1) as a main uptake route besides passive diffusion. In this work, CTR1 expression, cisplatin accumulation and intracellular copper concentration was assessed for single cells revisiting the case of CTR1 in the context of acquired cisplatin resistance. The single-cell workflow designed for in vitro experiments enabled quantitative imaging at resolutions down to 1 µm by laser ablation-inductively coupled plasma-time-of-flight mass spectrometry (LA-ICP-TOFMS). Cisplatin-sensitive ovarian carcinoma cells A2780 as compared to the cisplatin-resistant subline A2780cis were investigated. Intracellular cisplatin and copper levels were absolutely quantified for thousands of individual cells, while for CTR1, relative differences of total CTR1 versus plasma membrane-bound CTR1 were determined. A markedly decreased intracellular cisplatin concentration accompanied by reduced copper concentrations was observed for single A2780cis cells, along with a distinctly reduced (total) CTR1 level as compared to the parental cell model. Interestingly, a significantly different proportion of plasma membrane-bound versus total CTR1 in untreated A2780 as compared to A2780cis cells was observed. This proportion changed in both models upon cisplatin exposure. Statistical analysis revealed a significant correlation between total and plasma membrane-bound CTR1 expression and cisplatin accumulation at the single-cell level in both A2780 and A2780cis cells. Thus, our study recapitulates the crosstalk of copper homeostasis and cisplatin uptake, and also indicates a complex interplay between subcellular CTR1 localization and cellular cisplatin accumulation as a driver for acquired resistance development.

A platinum(IV) prodrug strategy to overcome glutathione-based oxaliplatin resistance.

Clinical efficacy of oxaliplatin is frequently limited by severe adverse effects and therapy resistance. Acquired insensitivity to oxaliplatin is, at least in part, associated with elevated levels of glutathione (GSH). In this study we report on an oxaliplatin-based platinum(IV) prodrug, which releases L-buthionine-S,R-sulfoximine (BSO), an inhibitor of glutamate-cysteine ligase, the rate-limiting enzyme in GSH biosynthesis. Two complexes bearing either acetate (BSO-OxOAc) or an albumin-binding maleimide (BSO-OxMal) as second axial ligand were synthesized and characterized. The in vitro anticancer activity of BSO-OxOAc was massively reduced in comparison to oxaliplatin, proving its prodrug nature. Nevertheless, the markedly lower intracellular oxaliplatin uptake in resistant HCT116/OxR cells was widely overcome by BSO-OxOAc resulting in distinctly reduced resistance levels. Platinum accumulation in organs of a colorectal cancer mouse model revealed higher tumor selectivity of BSO-OxMal as compared to oxaliplatin. This corresponded with increased antitumor activity, resulting in significantly enhanced overall survival. BSO-OxMal-treated tumors exhibited reduced GSH levels, proliferative activity and enhanced DNA damage (pH2AX) compared to oxaliplatin. Conversely, pH2AX staining especially in kidney cells was distinctly increased by oxaliplatin but not by BSO-OxMal. Taken together, our data provide compelling evidence for enhanced tumor specificity of the oxaliplatin(IV)/BSO prodrug.

Molecular and Pharmacological Bladder Cancer Therapy Screening: Discovery of Clofarabine as a Highly Active Compound.

BACKGROUND: The heterogeneity of bladder cancers (BCs) is a major challenge for the development of novel therapies. However, given the high rates of recurrence and/or treatment failure, the identification of effective therapeutic strategies is an urgent clinical need. OBJECTIVE: We aimed to establish a model system for drug identification/repurposing in order to identify novel therapies for the treatment of BC. DESIGN, SETTING, AND PARTICIPANTS: A collection of commercially available BC cell lines (n = 32) was comprehensively characterized. A panel of 23 cell lines, representing a broad spectrum of BC, was selected to perform a high-throughput drug screen. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Positive hits were defined as compounds giving >50% inhibition in at least one BC cell line. RESULTS AND LIMITATIONS: Amongst >1700 tested chemical compounds, a total of 471 substances exhibited antineoplastic effects. Clofarabine, an antimetabolite drug used as third-line treatment for childhood acute lymphoblastic leukaemia, was amongst the limited number of drugs with inhibitory effects on cell lines of all intrinsic subtypes. We, thus, reassessed the substance and confirmed its inhibitory effects on commercially available cell lines and patient-derived cell cultures representing various disease stages, intrinsic subtypes, and histologic variants. To verify these effects in vivo, a patient-derived cell xenograft model for urothelial carcinoma (UC) was used. Well-tolerated doses of clofarabine induced complete remission in all treated animals (n = 12) suffering from both early- and late-stage disease. We further took advantage of another patient-derived cell xenograft model originating from the rare disease entity sarcomatoid carcinoma (SaC). Similarly to UC xenograft mice, clofarabine induced subcomplete to complete tumour remissions in all treated animals (n = 8). CONCLUSIONS: The potent effects of clofarabine in vitro and in vivo suggest that our findings may be of high clinical relevance. Clinical trials are needed to assess the value of clofarabine in improving BC patient care. PATIENT SUMMARY: We used commercially available cell lines for the identification of novel drugs for the treatment of bladder cancer. We confirmed the effects of one of these drugs, clofarabine, in patient-derived cell lines and two different mouse models, thereby demonstrating a potential clinical relevance of this substance in bladder cancer treatment.

Development of a cobalt(iii)-based ponatinib prodrug system.

Receptor tyrosine kinase inhibitors have become a central part of modern targeted cancer therapy. However, their curative potential is distinctly limited by both rapid resistance development and severe adverse effects. Consequently, tumor-specific drug activation based on prodrug designs, exploiting tumor-specific properties such as hypoxic oxygen conditions, is a feasible strategy to widen the therapeutic window. After proof-of-principal molecular docking studies, we have synthesized two cobalt(iii) complexes using a derivative of the clinically approved Abelson (ABL) kinase and fibroblast growth factor receptor (FGFR) inhibitor ponatinib. Acetylacetone (acac) or methylacetylacetone (Meacac) have been used as ancillary ligands to modulate the reduction potential. The ponatinib derivative, characterized by an ethylenediamine moiety instead of the piperazine ring, exhibited comparable cell-free target kinase inhibition potency. Hypoxia-dependent release of the ligand from the cobalt(iii) complexes was proven by changed fluorescence properties, enhanced downstream signaling inhibition and increased in vitro anticancer activity in BCR-ABL- and FGFR-driven cancer models. Respective tumor-inhibiting in vivo effects in the BCR-ABL-driven K-562 leukemia model were restricted to the cobalt(iii) complex with the higher reduction potential and confirmed in a FGFR-driven urothelial carcinoma xenograft model. Summarizing, we here present for the first time hypoxia-activatable prodrugs of the clinically approved tyrosine kinase inhibitor ponatinib and a correlation of the in vivo activity with their reduction potential.

Genome-wide CRISPR screen identifies host dependency factors for influenza A virus infection.

Host dependency factors that are required for influenza A virus infection may serve as therapeutic targets as the virus is less likely to bypass them under drug-mediated selection pressure. Previous attempts to identify host factors have produced largely divergent results, with few overlapping hits across different studies. Here, we perform a genome-wide CRISPR/Cas9 screen and devise a new approach, meta-analysis by information content (MAIC) to systematically combine our results with prior evidence for influenza host factors. MAIC out-performs other meta-analysis methods when using our CRISPR screen as validation data. We validate the host factors, WDR7, CCDC115 and TMEM199, demonstrating that these genes are essential for viral entry and regulation of V-type ATPase assembly. We also find that CMTR1, a human mRNA cap methyltransferase, is required for efficient viral cap snatching and regulation of a cell autonomous immune response, and provides synergistic protection with the influenza endonuclease inhibitor Xofluza.

Resolving outbreak dynamics using approximate Bayesian computation for stochastic birth-death models.

Earlier research has suggested that approximate Bayesian computation (ABC) makes it possible to fit simulator-based intractable birth-death models to investigate communicable disease outbreak dynamics with accuracy comparable to that of exact Bayesian methods. However, recent findings have indicated that key parameters, such as the reproductive number R, may remain poorly identifiable with these models. Here we show that this identifiability issue can be resolved by taking into account disease-specific characteristics of the transmission process in closer detail. Using tuberculosis (TB) in the San Francisco Bay area as a case study, we consider a model that generates genotype data from a mixture of three stochastic processes, each with its own distinct dynamics and clear epidemiological interpretation.       We show that our model allows for accurate posterior inferences about outbreak dynamics from aggregated annual case data with genotype information. As a byproduct of the inference, the model provides an estimate of the infectious population size at the time the data were collected. The acquired estimate is approximately two orders of magnitude smaller than assumed in earlier related studies, and it is much better aligned with epidemiological knowledge about active TB prevalence. Similarly, the reproductive number R related to the primary underlying transmission process is estimated to be nearly three times larger than previous estimates, which has a substantial impact on the interpretation of the fitted outbreak model.

Likelihood-free inference via classification.

Increasingly complex generative models are being used across disciplines as they allow for realistic characterization of data, but a common difficulty with them is the prohibitively large computational cost to evaluate the likelihood function and thus to perform likelihood-based statistical inference. A likelihood-free inference framework has emerged where the parameters are identified by finding values that yield simulated data resembling the observed data. While widely applicable, a major difficulty in this framework is how to measure the discrepancy between the simulated and observed data. Transforming the original problem into a problem of classifying the data into simulated versus observed, we find that classification accuracy can be used to assess the discrepancy. The complete arsenal of classification methods becomes thereby available for inference of intractable generative models. We validate our approach using theory and simulations for both point estimation and Bayesian inference, and demonstrate its use on real data by inferring an individual-based epidemiological model for bacterial infections in child care centers.

Weak Epistasis May Drive Adaptation in Recombining Bacteria.

The impact of epistasis on the evolution of multi-locus traits depends on recombination. While sexually reproducing eukaryotes recombine so frequently that epistasis between polymorphisms is not considered to play a large role in short-term adaptation, many bacteria also recombine, some to the degree that their populations are described as "panmictic" or "freely recombining." However, whether this recombination is sufficient to limit the ability of selection to act on epistatic contributions to fitness is unknown. We quantify homologous recombination in five bacterial pathogens and use these parameter estimates in a multilocus model of bacterial evolution with additive and epistatic effects. We find that even for highly recombining species (e.g., Streptococcus pneumoniae or Helicobacter pylori), selection on weak interactions between distant mutations is nearly as efficient as for an asexual species, likely because homologous recombination typically transfers only short segments. However, for strong epistasis, bacterial recombination accelerates selection, with the dynamics dependent on the amount of recombination and the number of loci. Epistasis may thus play an important role in both the short- and long-term adaptive evolution of bacteria, and, unlike in eukaryotes, is not limited to strong effect sizes, closely linked loci, or other conditions that limit the impact of recombination.