The relatively slow mutation rates of bacterial pathogens impose severe limitations on phylodynamic analysis of bacterial outbreaks. However, whole-genome sequencing may enable accurate inference of bacterial transmission dynamics in health-care settings. We simulated the epidemic dynamics of a Staphylococcus aureus lineage using a stochastic model with a hospital and community compartment connected by patient admission and discharge. We generated synthetic genomic sequences and performed Bayesian phylodynamic inference on a proportion of samples from each simulated outbreak. When samples are obtained from both compartments, hospital transmission rate ([Formula: see text]) and community transmission rate ([Formula: see text]) are accurately estimated, if [Formula: see text] is on the same scale as the discharge rate. If [Formula: see text] is substantially lower than the discharge rate, a robust quantification of within-hospital transmission dynamics is challenging. Excluding samples from the community resulted in a notable underestimation of [Formula: see text] when [Formula: see text]. When transmission was 'community-driven', but sampling was restricted to hospital cases only, estimates are closer to the true [Formula: see text], if hospital sampling proportion is known. Otherwise, [Formula: see text] estimates reflected the transmission dynamics within the community. When using genomic data to estimate bacterial transmission rates in a health-care setting, it is essential to take into account the surrounding community. Many infections related to nosocomial outbreaks will not be observed within the hospital due to fast discharge rates. In the absence of usable genomic data from the community, alternative estimates of community transmission rates from publicly available data should be incorporated. Transmission rate estimates from nosocomial genomes alone need to be interpreted with care.

BACKGROUND: Estimating the time since HIV infection (TSI) at population level is essential for tracking changes in the global HIV epidemic. Most methods for determining TSI give a binary classification of infections as recent or non-recent within a window of several months, and cannot assess the cumulative impact of an intervention. RESULTS: We developed a Random Forest Regression model, HIV-phyloTSI, which combines measures of within-host diversity and divergence to generate continuous TSI estimates directly from viral deep-sequencing data, with no need for additional variables. HIV-phyloTSI provides a continuous measure of TSI up to 9 years, with a mean absolute error of less than 12 months overall and less than 5 months for infections with a TSI of up to a year. It performs equally well for all major HIV subtypes based on data from African and European cohorts. CONCLUSIONS: We demonstrate how HIV-phyloTSI can be used for incidence estimates on a population level.

Machine learning is increasingly introduced into medical fields, yet there is limited evidence for its benefit over more commonly used statistical methods in epidemiological studies. We introduce an unsupervised machine learning framework for longitudinal features and evaluate it using sexual behaviour data from the last 20 years from over 3'700 participants in the Swiss HIV Cohort Study (SHCS). We use hierarchical clustering to find subgroups of men who have sex with men in the SHCS with similar sexual behaviour up to May 2017, and apply regression to test whether these clusters enhance predictions of sexual behaviour or sexually transmitted diseases (STIs) after May 2017 beyond what can be predicted with conventional parameters. We find that behavioural clusters enhance model performance according to likelihood ratio test, Akaike information criterion and area under the receiver operator characteristic curve for all outcomes studied, and according to Bayesian information criterion for five out of ten outcomes, with particularly good performance for predicting future sexual behaviour and recurrent STIs. We thus assess a methodology that can be used as an alternative means for creating exposure categories from longitudinal data in epidemiological models, and can contribute to the understanding of time-varying risk factors.

Over the last decade, syphilis diagnoses among men-who-have-sex-with-men (MSM) have strongly increased in Europe. Understanding the drivers of the ongoing epidemic may aid to curb transmissions. In order to identify the drivers of syphilis transmission in MSM in Switzerland between 2006 and 2017 as well as the effect of potential interventions, we set up an epidemiological model stratified by syphilis stage, HIV-diagnosis, and behavioral factors to account for syphilis infectiousness and risk for transmission. In the main model, we used 'reported non-steady partners' (nsP) as the main proxy for sexual risk. We parameterized the model using data from the Swiss HIV Cohort Study, Swiss Voluntary Counselling and Testing center, cross-sectional surveys among the Swiss MSM population, and published syphilis notifications from the Federal Office of Public Health. The main model reproduced the increase in syphilis diagnoses from 168 cases in 2006 to 418 cases in 2017. It estimated that between 2006 and 2017, MSM with HIV diagnosis had 45.9 times the median syphilis incidence of MSM without HIV diagnosis. Defining risk as condomless anal intercourse with nsP decreased model accuracy (sum of squared weighted residuals, 378.8 vs. 148.3). Counterfactual scenarios suggested that increasing screening of MSM without HIV diagnosis and with nsP from once every two years to twice per year may reduce syphilis incidence (at most 12.8% reduction by 2017). Whereas, increasing screening among MSM with HIV diagnosis and with nsP from once per year to twice per year may substantially reduce syphilis incidence over time (at least 63.5% reduction by 2017). The model suggests that reporting nsP regardless of condom use is suitable for risk stratification when modelling syphilis transmission. More frequent screening of MSM with HIV diagnosis, particularly those with nsP may aid to curb syphilis transmission.

Beta-lactam- and in particular carbapenem-resistant Enterobacteriaceae represent a major public health threat. Despite strong variation of resistance across geographical settings, there is limited understanding of the underlying drivers. To assess these drivers, we developed a transmission model of cephalosporin- and carbapenem-resistant Klebsiella pneumoniae. The model is parameterized using antibiotic consumption and demographic data from eleven European countries and fitted to the resistance rates for Klebsiella pneumoniae for these settings. The impact of potential drivers of resistance is then assessed in counterfactual analyses. Based on reported consumption data, the model could simultaneously fit the prevalence of extended-spectrum beta-lactamase-producing and carbapenem-resistant Klebsiella pneumoniae (ESBL and CRK) across eleven European countries over eleven years. The fit could explain the large between-country variability of resistance in terms of consumption patterns and fitted differences in hospital transmission rates. Based on this fit, a counterfactual analysis found that reducing nosocomial transmission and antibiotic consumption in the hospital had the strongest impact on ESBL and CRK prevalence. Antibiotic consumption in the community also affected ESBL prevalence but its relative impact was weaker than inpatient consumption. Finally, we used the model to estimate a moderate fitness cost of CRK and ESBL at the population level. This work highlights the disproportionate role of antibiotic consumption in the hospital and of nosocomial transmission for resistance in gram-negative bacteria at a European level. This indicates that infection control and antibiotic stewardship measures should play a major role in limiting resistance even at the national or regional level.

Although combination antiretroviral therapies seem to be effective at controlling HIV-1 infections regardless of the viral subtype, there is increasing evidence for subtype-specific drug resistance mutations. The order and rates at which resistance mutations accumulate in different subtypes also remain poorly understood. Most of this knowledge is derived from studies of subtype B genotypes, despite not being the most abundant subtype worldwide. Here, we present a methodology for the comparison of mutational networks in different HIV-1 subtypes, based on Hidden Conjunctive Bayesian Networks (H-CBN), a probabilistic model for inferring mutational networks from cross-sectional genotype data. We introduce a Monte Carlo sampling scheme for learning H-CBN models for a larger number of resistance mutations and develop a statistical test to assess differences in the inferred mutational networks between two groups. We apply this method to infer the temporal progression of mutations conferring resistance to the protease inhibitor lopinavir in a large cross-sectional cohort of HIV-1 subtype C genotypes from South Africa, as well as to a data set of subtype B genotypes obtained from the Stanford HIV Drug Resistance Database and the Swiss HIV Cohort Study. We find strong support for different initial mutational events in the protease, namely at residue 46 in subtype B and at residue 82 in subtype C. The inferred mutational networks for subtype B versus C are significantly different sharing only five constraints on the order of accumulating mutations with mutation at residue 54 as the parental event. The results also suggest that mutations can accumulate along various alternative paths within subtypes, as opposed to a unique total temporal ordering. Beyond HIV drug resistance, the statistical methodology is applicable more generally for the comparison of inferred mutational networks between any two groups.

A key parameter in epidemiological modeling which characterizes the spread of an infectious disease is the generation time, or more generally the distribution of infectiousness as a function of time since infection. There is increasing evidence supporting a prolonged viral shedding window for COVID-19, but the transmissibility in this phase is unclear. Based on this, we develop a generalized Susceptible-Exposed-Infected-Resistant (SEIR) model including an additional compartment of chronically infected individuals who can stay infectious for a longer duration than the reported generation time, but with infectivity reduced to varying degrees. Using the incidence and fatality data from different countries, we first show that such an assumption also yields a plausible model in explaining the data observed prior to the easing of the lockdown measures (relaxation). We then test the predictive power of this model for different durations and levels of prolonged infectiousness using the incidence data after the introduction of relaxation in Switzerland, and compare it with a model without the chronically infected population to represent the models conventionally used. We show that in case of a gradual easing on the lockdown measures, the predictions of the model including the chronically infected population vary considerably from those obtained under a model in which prolonged infectiousness is not taken into account. Although the existence of a chronically infected population still remains largely hypothetical, we believe that our results provide tentative evidence to consider a chronically infected population as an alternative modeling approach to better interpret the transmission dynamics of COVID-19.

The scale-up of antiretroviral therapy (ART) in South Africa substantially reduced AIDS-related deaths and new HIV infections. However, its success is threatened by the emergence of resistance to non-nucleoside reverse-transcriptase inhibitors (NNRTI). The MARISA (Modelling Antiretroviral drug Resistance In South Africa) model presented here aims at investigating the time trends and factors driving NNRTI resistance in South Africa. MARISA is a compartmental model that includes the key aspects of the local HIV epidemic: continuum of care, disease progression, and gender. The dynamics of NNRTI resistance emergence and transmission are then added to this framework. Model parameters are informed using data from HIV cohorts participating in the International epidemiology Databases to Evaluate AIDS (IeDEA) and literature estimates, or fitted to UNAIDS estimates. Using this novel approach of triangulating clinical and resistance data from various sources, MARISA reproduces the time trends of HIV in South Africa in 2005-2016, with a decrease in new infections, undiagnosed individuals, and AIDS-related deaths. MARISA captures the dynamics of the spread of NNRTI resistance: high levels of acquired drug resistance (ADR, in 83% of first-line treatment failures in 2016), and increasing transmitted drug resistance (TDR, in 8.1% of ART initiators in 2016). Simulation of counter-factual scenarios reflecting alternative public health policies shows that increasing treatment coverage would have resulted in fewer new infections and deaths, at the cost of higher TDR (11.6% in 2016 for doubling the treatment rate). Conversely, improving switching to second-line treatment would have led to lower TDR (6.5% in 2016 for doubling the switching rate) and fewer new infections and deaths. Implementing drug resistance testing would have had little impact. The rapid ART scale-up and inadequate switching to second-line treatment were the key drivers of the spread of NNRTI resistance in South Africa. However, even though some interventions could have substantially reduced the level of NNRTI resistance, no policy including NNRTI-based first line regimens could have prevented this spread. Thus, by combining epidemiological data on HIV in South Africa with biological data on resistance evolution, our modelling approach identified key factors driving NNRTI resistance, highlighting the need of alternative first-line regimens.

Phylodynamic models are widely used in infectious disease epidemiology to infer the dynamics and structure of pathogen populations. However, these models generally assume that individual hosts contact one another at random, ignoring the fact that many pathogens spread through highly structured contact networks. We present a new framework for phylodynamics on local contact networks based on pairwise epidemiological models that track the status of pairs of nodes in the network rather than just individuals. Shifting our focus from individuals to pairs leads naturally to coalescent models that describe how lineages move through networks and the rate at which lineages coalesce. These pairwise coalescent models not only consider how network structure directly shapes pathogen phylogenies, but also how the relationship between phylogenies and contact networks changes depending on epidemic dynamics and the fraction of infected hosts sampled. By considering pathogen phylogenies in a probabilistic framework, these coalescent models can also be used to estimate the statistical properties of contact networks directly from phylogenies using likelihood-based inference. We use this framework to explore how much information phylogenies retain about the underlying structure of contact networks and to infer the structure of a sexual contact network underlying a large HIV-1 sub-epidemic in Switzerland.

UNLABELLED: The continuous time conjunctive Bayesian network (CT-CBN) is a graphical model for analyzing the waiting time process of the accumulation of genetic changes (mutations). CT-CBN models have been successfully used in several biological applications such as HIV drug resistance development and genetic progression of cancer. However, current approaches for parameter estimation and network structure learning of CBNs can only deal with a small number of mutations (<20). Here, we address this limitation by presenting an efficient and accurate approximate inference algorithm using a Monte Carlo expectation-maximization algorithm based on importance sampling. The new method can now be used for a large number of mutations, up to one thousand, an increase by two orders of magnitude. In simulation studies, we present the accuracy as well as the running time efficiency of the new inference method and compare it with a MLE method, expectation-maximization, and discrete time CBN model, i.e. a first-order approximation of the CT-CBN model. We also study the application of the new model on HIV drug resistance datasets for the combination therapy with zidovudine plus lamivudine (AZT + 3TC) as well as under no treatment, both extracted from the Swiss HIV Cohort Study database. AVAILABILITY AND IMPLEMENTATION: The proposed method is implemented as an R package available at https://github.com/cbg-ethz/MC-CBN CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Treatment failure after therapy of pulmonary tuberculosis (TB) infections is an important challenge, especially when it coincides with de novo emergence of multi-drug-resistant TB (MDR-TB). We seek to explore possible causes why MDR-TB has been found to occur much more often in patients with a history of previous treatment. We develop a mathematical model of the replication of Mycobacterium tuberculosis within a patient reflecting the compartments of macrophages, granulomas, and open cavities as well as parameterizing the effects of drugs on the pathogen dynamics in these compartments. We use this model to study the influence of patient adherence to therapy and of common retreatment regimens on treatment outcome. As expected, the simulations show that treatment success increases with increasing adherence. However, treatment occasionally fails even under perfect adherence due to interpatient variability in pharmacological parameters. The risk of generating MDR de novo is highest between 40% and 80% adherence. Importantly, our simulations highlight the double-edged effect of retreatment: On the one hand, the recommended retreatment regimen increases the overall success rate compared to re-treating with the initial regimen. On the other hand, it increases the probability to accumulate more resistant genotypes. We conclude that treatment adherence is a key factor for a positive outcome, and that screening for resistant strains is advisable after treatment failure or relapse.

Contact structure is believed to have a large impact on epidemic spreading and consequently using networks to model such contact structure continues to gain interest in epidemiology. However, detailed knowledge of the exact contact structure underlying real epidemics is limited. Here we address the question whether the structure of the contact network leaves a detectable genetic fingerprint in the pathogen population. To this end we compare phylogenies generated by disease outbreaks in simulated populations with different types of contact networks. We find that the shape of these phylogenies strongly depends on contact structure. In particular, measures of tree imbalance allow us to quantify to what extent the contact structure underlying an epidemic deviates from a null model contact network and illustrate this in the case of random mixing. Using a phylogeny from the Swiss HIV epidemic, we show that this epidemic has a significantly more unbalanced tree than would be expected from random mixing.

Human immunodeficiency virus (HIV-1) is, like most pathogens, under selective pressure to escape the immune system of its host. In particular, HIV-1 can avoid recognition by cytotoxic T lymphocytes (CTLs) by altering the binding affinity of viral peptides to human leukocyte antigen (HLA) molecules, the role of which is to present those peptides to the immune system. It is generally assumed that HLA escape mutations carry a replicative fitness cost, but these costs have not been quantified. In this study, we assess the replicative cost of mutations which are likely to escape presentation by HLA molecules in the region of HIV-1 protease and reverse transcriptase. Specifically, we combine computational approaches for prediction of in vitro replicative fitness and peptide binding affinity to HLA molecules. We find that mutations which impair binding to HLA-A molecules tend to have lower in vitro replicative fitness than mutations which do not impair binding to HLA-A molecules, suggesting that HLA-A escape mutations carry higher fitness costs than non-escape mutations. We argue that the association between fitness and HLA-A binding impairment is probably due to an intrinsic cost of escape from HLA-A molecules, and these costs are particularly strong for HLA-A alleles associated with efficient virus control. Counter-intuitively, we do not observe a significant effect in the case of HLA-B, but, as discussed, this does not argue against the relevance of HLA-B in virus control. Overall, this article points to the intriguing possibility that HLA-A molecules preferentially target more conserved regions of HIV-1, emphasizing the importance of HLA-A genes in the evolution of HIV-1 and RNA viruses in general.

Antibiotic resistant nosocomial infections are an important cause of mortality and morbidity in hospitals. Antibiotic cycling has been proposed to contain this spread by a coordinated use of different antibiotics. Theoretical work, however, suggests that often the random deployment of drugs ("mixing") might be the better strategy. We use an epidemiological model for a single hospital ward in order to assess the performance of cycling strategies which take into account the frequency of antibiotic resistance in the hospital ward. We assume that information on resistance frequencies stems from microbiological tests, which are performed in order to optimize individual therapy. Thus the strategy proposed here represents an optimization at population-level, which comes as a free byproduct of optimizing treatment at the individual level. We find that in most cases such an informed switching strategy outperforms both periodic cycling and mixing, despite the fact that information on the frequency of resistance is derived only from a small sub-population of patients. Furthermore we show that the success of this strategy is essentially a stochastic phenomenon taking advantage of the small population sizes in hospital wards. We find that the performance of an informed switching strategy can be improved substantially if information on resistance tests is integrated over a period of one to two weeks. Finally we argue that our findings are robust against a (moderate) preexistence of doubly resistant strains and against transmission via environmental reservoirs. Overall, our results suggest that switching between different antibiotics might be a valuable strategy in small patient populations, if the switching strategies take the frequencies of resistance alleles into account.

Despite comparable levels of virus replication, simian immunodeficiency viruses (SIV) infection is non-pathogenic in natural hosts, such as sooty mangabeys (SM), whereas it is pathogenic in non-natural hosts, such as rhesus macaques (RM). Comparative studies of pathogenic and non-pathogenic SIV infection can thus shed light on the role of specific factors in SIV pathogenesis. Here, we determine the impact of target-cell limitation, CD8+ T cells, and Natural Killer (NK) cells on virus replication in the early SIV infection. To this end, we fit previously published data of experimental SIV infections in SMs and RMs with mathematical models incorporating these factors and assess to what extent the inclusion of individual factors determines the quality of the fits. We find that for both rhesus macaques and sooty mangabeys, target-cell limitation alone cannot explain the control of early virus replication, whereas including CD8+ T cells into the models significantly improves the fits. By contrast, including NK cells does only significantly improve the fits in SMs. These findings have important implications for our understanding of SIV pathogenesis as they suggest that the level of early CD8+ T cell responses is not the key difference between pathogenic and non-pathogenic SIV infection.

Most population genetic theories on the evolution of sex or recombination are based on fairly restrictive assumptions about the nature of the underlying fitness landscapes. Here we use computer simulations to study the evolution of sex on fitness landscapes with different degrees of complexity and epistasis. We evaluate predictors of the evolution of sex, which are derived from the conditions established in the population genetic literature for the evolution of sex on simpler fitness landscapes. These predictors are based on quantities such as the variance of Hamming distance, mean fitness, additive genetic variance, and epistasis. We show that for complex fitness landscapes all the predictors generally perform poorly. Interestingly, while the simplest predictor, Delta Var(HD), also suffers from a lack of accuracy, it turns out to be the most robust across different types of fitness landscapes. Delta Var(HD) is based on the change in Hamming distance variance induced by recombination and thus does not require individual fitness measurements. The presence of loci that are not under selection can, however, severely diminish predictor accuracy. Our study thus highlights the difficulty of establishing reliable criteria for the evolution of sex on complex fitness landscapes and illustrates the challenge for both theoretical and experimental research on the origin and maintenance of sexual reproduction.