Pathogen Genomics in Public Health

Abstract

Rapid advances in DNA sequencing technology (“next-generation sequencing”) have inspired optimism about the future potential of human genomics for “precision medicine.” Meanwhile, pathogen genomics is already delivering “precision public health” via more effective foodborne illness outbreak investigations, better targeted tuberculosis control, and more timely and granular influenza surveillance to inform vaccine strain selection. In this article, we describe how public health agencies are rapidly adopting pathogen genomics to improve their effectiveness in almost all domains of infectious disease. This momentum is likely to continue, given ongoing development in sequencing and sequencing-related technologies.

An important transformation is underway in public health. Next-generation sequencing (NGS, also called “high-throughput sequencing”) is reshaping communicable disease surveillance, allowing for earlier detection and more precise investigation of outbreaks. NGS helps characterize microbes more effectively and offers new insights into their ecology and transmission. The plethora of sequence data provides raw material for research and development of new diagnostics and therapeutics.

This article describes how pathogen genomics has been changing public health in the United States and globally.

Adapting NGS to Public Health Use

The NGS era began with the commercial release of massively parallel pyrosequencing in 2005, the first fundamental advance in sequencing technology since the invention of Sanger sequencing in the 1970s. ^1,2 In the early years, NGS efficiency improved rapidly, with sequencing costs falling by as much as 80% year-over-year. ^1,3 In public health, these developments were both exciting—because of the myriad potential applications, including bacterial whole-genome sequencing (WGS)⁴—and intimidating—because of the barriers: implementing NGS would require investment in sequencing equipment as well as high-performance computing infrastructure to move, store and analyze large volumes of sequence data. Equally important was the need to integrate bioinformatics, a discipline new to public health.

Public Health England was an early leader in the use of NGS at a national scale, particularly for tuberculosis^5,6 and bacterial foodborne disease surveillance.^7,8 In the United States, CDC was a late adopter,⁹ but is now applying the technology broadly, due largely to the Advanced Molecular Detection (AMD) program, a $30 million dollar per year initiative established by Congress in 2013 to bring NGS and other innovative laboratory technologies to bear against infectious disease threats, first at CDC and then in state and local public health departments nationwide.

Applications of Pathogen Genomics

Today, pathogen genomics is part of almost every infectious disease program at CDC.¹⁰ Some applications of NGS that serve specialized purposes, such as reference testing, are in use only at CDC, while others drive entire domestic surveillance systems. Below, we provide examples to highlight the value of NGS technology for public health (Box).

Box: Generalizations about Sequencing in Public Health.

Several attributes of next-generation sequencing are driving adoption of the technology within public health:

• High resolution subtyping of pathogens

  • Examples

    • Bacterial enteric illness: improves detection of and response to outbreaks.
    • Tuberculosis: allows better targeting of interventions to stop transmission.
    • Legionella: provides new tool to understand the ecology of the pathogen in water systems.
    • Potential agents of bioterrorism: allows for improved forensics.
  • Caveats: Legacy technologies often need to be continued during a transition period, since older subtyping characterizations often can’t be reliably predicted from nucleic acid sequence data; PFGE patterns, for example, usually cannot be predicted from routine whole-genome sequencing.
• Efficient inference of phenotypic traits

  • Examples

    • Serotyping: In US public health labs, influenza viruses are now subject to a “sequencing first” approach, in which antigenic type and subtype can be inferred from the sequence; only a subset of viruses undergo traditional typing and subtyping. For pathogens such as E. coli,^15,18,24 Salmonella^25,26 or pneumococcus,⁹⁵ serotype can usually be inferred from sequence data, without the need to acquire and maintain serum panels.
    • Antimicrobial resistance: for bacterial pathogens such as Salmonella,^21,22 Escherichia coli,²³ Streptococcus,^75,95 MTB, or gonococcus⁹⁶ (to name but a few), antimicrobial resistance is increasingly inferred from genomic data.
    • Virulence: known virulence factors, such as the presence or absence of Shiga toxin genes in an E. coli strain, can also be inferred from genomic data.
  • Caveats: There will probably always be a need for traditional phenotyping. The ability to predict a phenotype from a genome generally relies on known correlations between the phenotypic characteristics and specific genetic sequences. Particularly in rapidly evolving species such as influenza, those correlations will need constant updating. In addition, the consistency of those correlations is variable. The reliability of inferred antimicrobial resistance, for example, is highly dependent on the type of antibiotic, the mechanism of resistance, and the species of bacteria. This reliability should improve over time as more data become available and algorithms for predicting phenotype improve. The capability to infer phenotype from genotype means that fewer traditional tests will need to be done in the future, and that fewer laboratories (i.e., reference laboratories) will need to maintain the capacity to perform them.
• “Deep sequencing” rather than “consensus sequencing”. Whereas Sanger sequencing generally provides a single, “consensus” sequence from a sample, NGS typically provides many (often hundreds, thousands or more) “reads” of the gene or amplicon.

  • Examples

    • Malaria: In highly endemic areas, infection with multiple strains of malaria is common. In such cases, Sanger sequencing usually reflects only the most dominant strain in the individual and can miss the presence of other strains, which may have differing resistance to anti-malarial agents. In malaria endemic areas, deep sequencing can also be used to quantify the number of strains in an individual, a correlate of the intensity of transmission and potentially a tool for evaluating the impact of community interventions.
    • Hepatitis C: Hepatitis C virus mutates rapidly in individuals, resulting in a “swarm of quasispecies”. Data on the diversity of quasispecies in two individuals provides a reliable means of inferring whether they are part of a single outbreak.^97,98
    • Influenza: High mutation rates in influenza virus can also lead to minor variants with resistance to oseltamivir or other antivirals, which could be missed by consensus sequencing.
  • Caveats: Sequencing errors, which are more common with NGS than with Sanger sequencing, can create the illusion of rare variants. Careful analysis of potential sequence variants is needed to prevent this.
• More efficient workflows. Characterizing pathogens by means of sequencing is sometimes but not universally less laborious and less expensive than traditional typing.

  • Examples

    • Escherichia coli: Shiga toxin-producing E coli (STEC) are a common cause of foodborne illness. A standard workflow for characterizing STEC involves determination of serotype, traditionally accomplished by means of a panel of antisera to determine O antigen and, if indicated, H antigen, as well as detection of Shiga toxin or Shiga toxin genes by, for example, PCR. All of these characteristics, as well as susceptibility or resistance to several antibiotics, can be reliably inferred from whole-genome sequencing.
    • Influenza virus: The United States is now using a “sequence first” approach to influenza virus characterization (see text).

• Caveats: Sequencing, for the time being, is often more expensive than traditional subtyping alone. For bacterial pathogens, for example, whole-genome sequencing is typically twice as expensive as PFGE alone. However, if PFGE for a particular pathogen needs to be accompanied by traditional phenotyping such as serotyping, virulence typing, or antimicrobial resistance testing, and those features can be reliably inferred from the genome, then the sequencing approach is often less expensive.

Bacterial foodborne illness.

In the mid-1990s, U.S. foodborne disease programs first began applying standardized molecular subtyping—pulsed-field gel electrophoresis (PFGE)—to bacterial pathogens as part of routine surveillance, leading to a fundamental change in how outbreaks were identified and investigated. The resulting national network, “PulseNet”, now includes more than 80 public health laboratories.¹¹ Before PulseNet, outbreaks were difficult to detect and solve unless they were large or very geographically and temporally focused. For example, during the 20-year period before PulseNet, only five outbreaks of listeriosis (0.25 per year) were solved (i.e., with a food source identified),¹² with a mean of 54 cases per outbreak. In the 5-year period after PulseNet began, 11 outbreaks were identified (2.2 per year) with a median of 5 cases per outbreak.¹² Routine use of PFGE had a similar impact on the detection and response to other foodborne bacteria, particularly Salmonella¹³ and Shiga toxin-producing Escherichia coli (STEC).^14–16

PulseNet has now transitioned from PFGE to WGS.^17,18 Partners in this effort include the U.S. Food and Drug Administration, whose GenomeTrakr system^19,20 performs WGS of food and environmental isolates, the U.S. Department of Agriculture, and the National Center for Biotechnology Information (NCBI).

Compared with PFGE, WGS offers a vastly finer resolution: typically, a three- to six-million base-pair sequence, in contrast to a gel pattern with ten to twenty bands that reflect changes in small parts of the genome. WGS data are inherently digital, standardized, and much less dependent on the choice of laboratory protocol. The results reveal evolutionary relationships between bacterial isolates, allowing a better understanding of transmission and links between cases (Figure 1). WGS can also predict phenotypic characteristics, such as virulence, serotype and antimicrobial resistance.^15,21–26 Costs for WGS (around $200 to $250 per isolate, including consumables, labor, equipment, maintenance and overheads) are currently still higher than those for PFGE (around $100 according to the same analysis), although the higher costs may be partly or entirely offset by eliminating the need for traditional phenotyping assays. In addition, advances in sequencing technology and laboratory automation may further reduce WGS costs.

Figure 1: Example of sequencing for outbreak detection and investigation.

An important purpose of infectious disease surveillance is identifying outbreaks for investigation and intervention. Discovering patterns in the epidemiologic data—i.e., finding common exposures among cases that cluster in time and location—can help distinguish outbreaks from the often much larger background of sporadic cases. Molecular subtyping has played an increasingly central role in this process by detecting cases with isolates that share a common molecular “fingerprint.”

In this figure, we schematically represent surveillance data for a foodborne pathogen, Salmonella enterica serovar Enteritidis, reported from one region of the United States in 2018; in that year, some states in the region were already sequencing Salmonella isolates in real time and others had not yet started. In the three panels, each dot represents a case of Salmonella Enteritidis gastroenteritis. Gray dots represent cases that were later determined to be “sporadic” (i.e., not linked to outbreaks) and colored dots represent cases that were eventually linked to outbreaks. The largest of these outbreaks (red dots) began as two distinct clusters of disease associated with restaurants in two different states. Whole-genome sequencing (WGS) linked these two clusters together and to several other cases outside the region.

The first panel (“Unsorted”), displays cases randomly, without regard to molecular subtyping.

The second panel represents a grouping of cases based on results of pulsed-field gel electrophoresis (PFGE), a molecular subtyping technology that US public health agencies have used since the 1990s. In this example, PFGE was mostly successful at grouping cases from the largest (red) outbreak; however, the group includes many cases unrelated to the outbreak, complicating the investigation and reducing the likelihood of finding the food source.

In the third panel, we show that the finer resolution afforded by whole-genome sequencing (WGS) was more effective in segregating the red outbreak cases from others. This gave investigators more confidence in the cluster definition and allowed them to focus on cases that were more likely part of the same outbreak. In this case, epidemiologic investigation identified shell eggs as the likely source, which was quickly confirmed by isolating Salmonella Enteritidis from the implicated eggs and confirming that its WGS matched that from the outbreak cases.

In addition to the outbreak in red, this panel shows four additional outbreaks. Cases in blue were part of a restaurant-associated outbreak linked to chicken in a single state. Two cases (green) were linked to live poultry exposure as part of a much larger, multistate, multi-strain outbreak that occurred mostly outside the region shown here. The five cases in light pink were investigated as an outbreak but no food source was identified. The fifteen cases in light orange occurred in a state where real-time WGS had not yet been implemented; their isolates were not sequenced until a later date, after the apparent outbreak had ended. This figure summarizes relationships identified by WGS using a simplified graph; in practice, however, the data would be represented as a phylogenetic tree, which contains additional detail that more precisely represents the relationships among sequences.

It is too early to know how the transition to WGS will affect U.S. surveillance for more common foodborne pathogens, such as Salmonella and STEC; however, early experience with Listeria surveillance, which switched to routine WGS in 2014, has been encouraging. In the first three years of WGS (September 2013 through August 2016), 18 outbreaks of listeriosis were solved (6 per year) with a median of 4 cases per outbreak.²⁷ In the United Kingdom, where WGS has been in routine use for STEC since at least 2015, the number of clusters detected has doubled.²⁸

Tuberculosis.

Since the 1990s, several DNA fingerprinting technologies have proven useful for subtyping Mycobacterium tuberculosis (MTB).²⁹ Identifying closely-related strains allows health department investigators to detect clusters of cases that may be linked to recent transmission—cases requiring more intense investigation and possible intervention.³⁰ WGS offers much finer resolution subtyping than older technologies and thus more confidence in the inferred relationships among cases. After using WGS selectively for several years, the U.S. tuberculosis control program has now scaled up to sequence isolates from all culture-confirmed cases nationwide. In California, WGS allowed public health workers to refute more than half of suspected outbreaks initially identified by conventional genotyping, saving time and resources (personal communication, Tambi Shaw, California Department of Public Health). Early experience in U.K.,⁵ Canadian,^31,32 and Dutch³³ tuberculosis programs has also confirmed that WGS supports more effective investigations by more accurately defining outbreaks^5,31,33, providing insights into transmission dynamics,³⁴ and sometimes suggesting the presence of previously unidentified cases or possible “super-spreaders” that should be prioritized for isolation and treatment.^5,31 WGS may also indicate whether recurrent cases are due to reactivation or reinfection, information useful in evaluating program effectiveness.³⁵

The ability to prioritize case-investigations may also be of use in the high-incidence, low- and middle-income countries which carry the world’s heaviest burden of tuberculosis.³⁶ In these countries, however, a different application of NGS—sequencing of MTB directly from sputum—could have an even more important role.³⁷ Direct sequencing of MTB from smear-positive sputum samples is already feasible in research settings^38–40 but is too expensive and cumbersome for routine clinical and public health use. If it can be made practical and cost-effective, this approach will enable rapid inference of drug susceptibility, which is already quite accurate for most first-line drugs, and will improve as more data become available.^37,41 In addition to supporting prompt treatment with appropriate drugs, NGS will reduce the need for routine phenotypic testing, which is complex, slow, and difficult to maintain in resource-limited laboratory settings.

In the meantime, an intermediate strategy is already practical in high-income countries: WGS directly from early positive cultures, providing information on drug susceptibility weeks before the results of traditional tests are available.³⁸ Laboratories in both the New York State Department of Health and Public Health England³⁷ have received regulatory approval to forego traditional drug-susceptibility testing of isolates predicted by WGS to be susceptible to all four first-line drugs—approximately 70–80% of all isolates.³⁷

Another promising strategy is targeted amplicon sequencing of selected mycobacterial genes or marker sequences.⁴² To remain relevant over time, any sequence-based method for inferring drug susceptibility must rely on the continuous updating of databases with correlated genotypic and phenotypic data.³⁷

Influenza.

The selection of seasonal influenza vaccine candidate strains is a complex, global undertaking, involving massive surveillance efforts from dozens of countries and contributing organizations. The World Health Organization (WHO) convenes international experts twice yearly to review information on circulating influenza strains and, based on that information, to recommend components of the Northern and Southern Hemisphere influenza vaccines.⁴³ CDC contributes to this process by overseeing characterization of 4,000 to 10,000 influenza specimens each year.⁴⁴

The traditional method for characterizing influenza strains begins with viral culture, which is increasingly challenging for certain strains, particularly H3N2.⁴⁵ Two or more passages through culture are often required, during which some adaptation of the virus may occur. Next, a few viral isolates are selected for phenotyping, which previously included antigenic characterization and consensus (Sanger) sequencing of selected genes. These steps are time-consuming and labor-intensive.

NGS now enables a more efficient “sequence-first” approach, in which original specimens are subjected directly to whole-genome reverse-transcriptase PCR, followed by sequencing.^45–47 These sequence data provide a highly granular view of viral emergence and allow for a more parsimonious selection of viruses for phenotypic characterization, including antigenic analysis and susceptibility to antiviral agents. This approach is not only faster but also more informative. For example, detailed phylogenetic information on all viral segments provides a richer picture of how influenza viruses are diversifying to evade existing immunity, and deep sequencing (sequencing many copies of the genome from the same sample) can reveal the presence of drug-resistant minor variants not reflected in the consensus sequence. Sequencing cannot completely replace traditional phenotyping, but sequencing first, using clinical specimens, allows it to be done more selectively.

Influenza NGS data are now routinely reviewed at the biannual WHO consultations and have already affected vaccine decision-making in at least two major instances, most recently contributing to a change in the A(H3N2) vaccine component to target a newly emerging clade.^45,48,49 NGS data are also used for forecasting the relative importance of emerging strains and risk assessment,^50–52 characterizing viruses used in vaccine effectiveness studies,⁵³ and informing treatment for patients infected with high pandemic risk viruses such as H7N9.^54,55

Parasitic diseases.

Diagnosis of many parasitic diseases continues to rely on microscopy, a nineteenth-century technology that is operator-dependent and resistant to automation. PCR and other diagnostic techniques (e.g., serology) have been developed for many common parasites but require separate tests for each suspected pathogen.

CDC’s parasitic diseases laboratories are developing a new type of diagnostic test based on the targeted and direct sequencing of eukaryotic housekeeping genes. This approach should enable the accurate detection of all known potential parasitic agents present in a blood sample with a single test. Early validation data suggest that this novel assay is at least as sensitive as standard PCR for parasites found in blood.⁵⁶ Further validation and development is ongoing, with plans to add more targets and to adapt the assay to more complex samples, including tissue and stool.

PCR amplification and NGS of specific genes is an effective means of identifying drug-resistance in malaria parasites.⁵⁷ Because such testing takes 2–3 days and is still somewhat expensive, its use in routine patient care remains limited. For surveillance at the country or regional level, however, it is a cost-effective and efficient way to assess drug resistance and to target treatment recommendations more precisely. NGS-based protocols may also be useful for assessing the intensity of transmission by gauging multiplicity of infection among residents in malaria-endemic areas.⁵⁸

Cyclospora cayetanensis causes outbreaks of foodborne diarrheal disease in the US every year. The organism’s limited genotypic variability and its inability to be propagated in the laboratory have confounded the development of effective genotyping methods for surveillance. CDC has developed a method to extract C. cayetanensis directly from stool and used it to produce whole genomes of multiple isolates.⁵⁹ Several promising genotyping targets have been identified and their translation into a functional and discriminatory genotyping tool is showing promise.⁶⁰ The genotyping tool and associated analysis system underwent further testing and validation during 2018’s unusually active summer cyclosporiasis season.

Other applications.

NGS is applicable across the spectrum of important pathogens in public health. For Legionnaires’ disease, for example, finer subtyping has been useful for investigating and responding to outbreaks.^61,62 Eventually, insights from NGS into Legionella ecology and persistence in water systems could improve prevention.⁶³ For hospital acquired infections, NGS is proving to be an invaluable tool for identifying and investigating outbreaks^64,65 and also for better understanding transmission both at the hospital⁶⁵ and community^64,66 level. For HIV, genetic sequence data generated for clinical purposes can be analyzed to identify potential clusters for early public-health intervention;^67,68 user-friendly tools⁶⁹ now allow state and local health departments to make use of these data. Community-level molecular surveillance for hepatitis C clusters has also proven useful.⁷⁰

Other applications for NGS in public health include tracking the emergence of antimicrobial resistance and novel resistant pathogens such as Candida auris,^66,71 tracking insecticide resistance in mosquito vectors of disease,^72,73 monitoring streptococcal pathogens,^74,75 investigating potential clusters of meningitis,⁷⁶ and many more.

Sequencing to support outbreak response.

An important subset of NGS applications in public health involves outbreak response.⁷⁷ For bacterial foodborne disease, for example, sequencing is central to detecting outbreaks, investigating cases, confirming the implicated food and tracing it back to its source. A very different example comes from the latter phases of the 2014-’16 Ebola outbreak in Guinea, in which sequencing was useful in ascertaining the likely source of infection in “outlier” cases—those with no known connection to other cases. Each outlier raised troubling questions: Did the case represent another introduction from an animal reservoir? Was it attributable to sexual transmission? Had a long chain of transmission been missed, suggesting serious gaps in surveillance? Each of these would require a different response. Fortunately, in this outbreak, sequences from outliers were consistently closely related to those in the known outbreak zone, and the response team was able to remain focused on stopping transmission there.^10,78

In contrast, sequencing Zika virus during its emergence in the Americas was not useful for responding to individual cases. However, sequence data were critical for developing diagnostics and vaccines, to better understand the evolution of the epidemic,⁷⁹ and recently to support evidence of an undetected outbreak in Cuba.⁸⁰ Analysis of Ebola virus genomes after the West Africa outbreak also provided useful insights into the virus’s spread.⁸¹

Where to Go Next?

Six years into the AMD program, NGS is now central to US public health programs for monitoring, controlling, and preventing infectious diseases. Progress is currently needed in several areas:

Metagenomics.

Sequencing, particularly bacterial WGS, often requires pure cultures of the organisms, which are increasingly difficult to obtain as clinical laboratories expand the use of highly multiplexed, syndrome-based, culture-independent diagnostic tests.^82,83 Most immediately, this trend is adversely affecting enteric disease surveillance as clinical laboratories move away from culture.⁸⁴ One solution may be to bypass culture, sequencing pathogen genomes directly from specimens.^38,85 Although already feasible for selected organisms in particular specimens, the methods are not yet practical for routine use.

Data integration and data science.

In public health settings, laboratory and epidemiologic data are often stored and managed separately. Until recently, these data could be brought together for analysis without loss of information: laboratory data such as a positive/negative result, a titer, or even a PFGE pattern identifier could be imported into an epidemiologic database and analyzed with traditional statistical tools. With pathogen genomic data, this is no longer true: these data need to be integrated to realize the full value of both.^31,86,87

Fortunately, academic research is addressing this challenge,⁸⁸ producing tools for visualizing and analyzing epidemiologic and phylogenetic data together, such as Microreact (microreact.org),⁸⁹ Nextstrain (nextstrain.org)⁹⁰ or the Interactive Tree of Life (itol.embl.de).⁹¹ More broadly, the emerging field of data science offers novel approaches for integrating, analyzing and visualizing increasingly diverse public health data.^20,92

Software to facilitate NGS workflows.

Complex workflows are required to manage the sequencing process, analyze raw sequence data, store processed data and integrate it with epidemiologic data, and finally, share information securely. In a single academic lab, it’s feasible for a bioinformatician to accomplish this, but in a network of public health laboratories, it can be more practical to manage much of this centrally. Bioinformatics tools are available to accomplish the basic steps of assembling raw sequence data into genomes and “pipelines” using these tools are available to automate core processes, such as validating quality, assembling a genome, and inferring phenotypes. However, user-friendly tools for managing workflows and integrating data are often lacking. In other instances, too many tools exist and it is not clear which ones will survive the test of time.

Challenges

Workforce transformation, both among microbiologists and epidemiologists, while perhaps the most obvious hurdle, is in some ways the least challenging due to enthusiasm about genomics. Even recruiting and retaining bioinformaticians has not been as difficult as initially anticipated. Among the first 27 recruits into the CDC/Association of Public Health Laboratories Bioinformatics Fellowship, 19 (70.4%) are still working in public health, despite a competitive marketplace. Those who stay in public health often cite the opportunity to have positive social impact as a key motivator.

Costs represent a more difficult issue. For now, sequencing is often more expensive than traditional subtyping. In addition, to minimize per-sample costs, sequencing technologies such as the MiSeq platform (Illumina, San Diego, Calif.) may require batch sizes of 15 samples or more, which can extend turn-around times beyond the 36 hours needed for the sequencing itself. Clearly, during a fast-moving outbreak, such delays are undesirable. Single-molecule, long-read sequencing technology, most notably the MinION (Oxford Nanopore Technologies, Oxford, UK), already has niche uses within public health because of its portability, but also has the potential to reduce batch-size needs and turn-around times.⁷⁸

Data Openness

In both academia and public health, pathogen genomics is ushering in a new era in data openness. In the United States, local, state and federal agencies are uploading data on bacterial foodborne pathogens,²⁰ influenza⁴⁵ and other pathogens to public databases hosted by NCBI (ncbi.nlm.nih.gov/pathogens), making these data available in near real time. These groups also contribute to other global databases, such as the Relational TB Sequencing Data Platform (“ReSeqTB”, reseqtb.org) and the Global Initiative on Sharing All Influenza Data (GISAID, gisaid.org), established to promote international exchange of sequence and other data. In addition to enabling secondary uses of data, openness also encourages collaboration among public health organizations, academia, and industry. Nevertheless, openness can never be complete and unconditional: public health agencies have always been vigilant guardians of confidentiality, and pathogen genomic data are released only after careful consideration of risks.^87,93,94

Conclusion

NGS and bioinformatics are transforming the response to infectious disease outbreaks, providing new insights into disease emergence and transmission, expediting pathogen characterization, and promoting data sharing (Figure 2). In public health, sequencing is already routine in many core domains, including foodborne bacterial pathogens, tuberculosis, influenza and antimicrobial resistance. These developments are taking place within a rapidly evolving technology landscape: NGS is becoming more automated, efficient, and accurate and related technologies, such as systems for highly multiplexed DNA amplification, are advancing.

Figure 2.

Typical Pathogen Genomics Workflow. From pathogens (a) collected in the course of disease surveillance, genomic DNA (b) or RNA are extracted, shorn into shorter segments, labeled, and subjected to next-generation (high-throughput) sequencing (c). The raw data from the sequencer are sorted, reassembled and aligned to other genomes for comparison (d). The assembled genomes are used for several purposes including determining relatedness (e) and predicting phenotypic traits such as virulence, antimicrobial resistance, and serotype (f). Increasingly, the data made publicly available in real-time for use by researchers and for the development of diagnostics, therapeutics and vaccines (g).

Public health workforce development is central to this process. Microbiologists need a strong knowledge base in microbial genomics. Epidemiologists need the skills and tools to translate genomic data into public-health action. Both groups need to grasp the basic vocabulary of bioinformatics. Public health should strive to attract professionals with broadly applicable data-science skills. For anyone considering a career in public health, this is an exciting time to jump in.