2023 APHL/CDC Advanced Molecular Detection (AMD) Days Conference
The Centers of Disease Control and Prevention (CDC) Office of Advanced Molecular Detection (OAMD) just celebrated its tenth anniversary. I attended the AMD Days Conference (12–14 September 2023) to learn more about intersection of genomics and public health.
What is in the scope of Advanced Molecular Detection?
TL;DR:
- Mostly infectious diseases
- Mostly next-generation sequencing (NGS)
- Mostly on pathogen genomes
I have observed about a dozen different places within CDC involved in genomics, so I wanted to answer the question “What is in scope— and what is out of scope — of OAMD?” What I found is that for every person I asked, I got a different answer. Each person’s assertion about in-scope and out-of-scope was violated by some aspect of content at the conference.
Let’s start with the tag-line: “Advanced molecular detection (AMD) integrates the latest next-generation genomic sequencing technologies with bioinformatics and epidemiology expertise across CDC and the nation to help us find, track, and stop disease-causing pathogens faster than ever before.”
| Hypothesis | Advanced | Molecular | Detection |
|------------|------------------------------------|--------------------------|--------------------------------|
| Tag-line | Next-generation sequencing (NGS) | Nucleic Acids (DNA, RNA) | Pathogen |
| 2 | | All Biomolecules | Infectious-diseases |
| 3 | | | Infectious + Newborn Screening |
| 4 | NGS + Mass Spec (MALDI-TOF, MS/MS) | | |A poster or presentation breaks each of these hypothesis. For example…
Poster 4–495, “Assessing the Diversity of Ebolavirus Glycoproteins in Outbreaks and the Use of Medical Countermeasures.” The authors use a high content imaging neutralization assay to see if an antibody cocktail still works against a mutated Ebola strain in vitro. Breaks NGS and DNA/RNA.
Poster 4–499, “Early Warning Signatures of Silicosis Revealed by Machine Learning and Gene Expression Profiles.”The authors looked at differential gene expression for three types of people: health/control, exposed, and diagnosed. They developed a model that could separate healthy people, but could not (yet) distinguish exposed from diagnosed. And the only way to detect it today is a pathologist manually looking at a biopsy. This is a non-infectious disease caused by a mineral, an environmental factor, and looks a human as opposed to pathogen DNA. Breaks everything but “uses DNA.”
The State Perspective
While CDC and APHL were well represented, most attendees came from state public health labs. I spoke in-depth to representatives of 10 states and 1 city.
State public health labs are highly heterogeneous
Some states have world class NGS core facilities that routinely publish in leading journals: MI, NY (Wadsworth), WA, WI. Other states do not have a single bioinformatician on staff; these usually rely on support from a local university. The median state has a single bioinformatician and/or a single genetic epidemiologist.
Salami-sliced funding
State Public Health Labs do not experience CDC funding — let alone Federal funding — as a monolith. Rather, dozens of CDC budget lines fund dozens of budget lines within the State public health lab.
The 1,000 lb gorilla of public health lab funding is the CDC Epidemiology and Laboratory Capacity (ELC) cooperative agreement. In FY2022, ELC dispersed $200M in core funding and $150M in COVID funding to 64 ELC jurisdictions (excludes tribes; includes all states, DC, all territories, and four large cities: Chicago, Houston, Los Angeles, New York City).
Lets take a microcosm of the salami slicing issue: New York City. For FY2022, there are six budget lines just for ELC [^1]. And the ELC budget lines change from year to year.
In addition to the ELC block, Public Health labs must
- Track funding across both pathogen-agnostic and pathogen-specific funding streams
- Account for compliance restrictions within each
- Satisfy reporting requirements. The reporting is not consolidated across the Federal government (e.g., CDC, DHS, FDA), nor is it consolidated across CDC funding streams. Meaning the agency must repeat itself in slightly different ways.
Examples Federal funding for Public Health Labs include…
- CDC NARMS ( Antimicrobial resistance)
- CDC HAI (Healthcare Associated Infections)
- CDC NWSS ( Wastewater )
- CDC FoodNet
- CDC PulseNet (DNA fingerprints)
- CDC SSuN (STD Surveillance Network)
- DHS BioWatch
- FDA GenomeTrakr
Some of these are reflected as a budget line (e.g., CDC SSuN and DHS BioWatch are reflected in the New York City Department of Health and Mental Hygiene budget), others are not — presumably included as a restriction or set-aside within another budget line.
Cannot hire informatics skills
State public health labs cannot hire bioinformaticians and genomic epidemiologists. Some states cannot hire because they do not have the “slots.” Other states have slots that go unfilled because the salaries are too low to attract qualified candidates.
Some states have responded with structured programs to upskill their existing (mostly laboratory) staff. Example: Poster, 1–496. “Workforce Development for Developing a Comprehensive Public Health Pathogen Genomics Program.” Where the author worked with a commercial company to up-skill staff.
Cannot retain lab skills
A recurring theme was that laboratory scientists feel forced out of the lab. There is a ceiling to advancement, so to make more money or accrue more prestige, one must leave the lab — moving into management or leaving public health entirely.
My ad hoc maturity model
I have observed that the OAMD maturity of a public health lab can be assessed with a relatively small set of questions:
- Do the public health lab and the epidemiologsts sit in the same government department: [Yes, No]
- Core Facility: [Yes, No]
- Are the bioinformaticians: [In-house, University, Commercial Contract]
- If bioinformatics are in-house, do they sit with: [Epidemiology, Lab , IT]
- Can the public health lab process clinical samples (CLIA, as opposed to surveillance and research): [Yes, No]
- To what degree is the lab automated: [bioinformatics pipelines, sequencing prep, interoperability with case report]
The features of the most mature public health labs are bold.
I find that the “business architecture” of the state public health lab is the second biggest determinant of capacity, behind funding. If the public health lab and the epidemiologsts are are different government departments, every problem of communication/interoperability between lab and case is exacerbated.
The Federal Perspective
Tension: Pushing the limits of advanced vs raising the bar
The “A” in AMD is Advanced. There are bleeding edge companies working on direct protein sequencing, no-amplification environmental detection, etc. Think real-life multi-modal Tricorders of Star Trek fame.
However, most states are struggling with 0 or 1 bioinformatician. And cannot run, let alone scale, what are considered routine assays from more mature states. From the Federal resource allocation perspective, this presents a tension.
Every dollar going towards pushing the envelope is a dollar not going to the real-world needs of public health departments right now. What good is yet another publication from Wadsworth, when New York City can barely keep up with Mpox testing now, and doesn’t have the capacity for tomorrow’s public health scare de jure.
There is also an equity impact. The “advanced” dollars go to the advanced states, leaving the median state even further behind.
The same problem plays out in software with funding highly configurable/customization tooling for researchers and bioinformaticians vs robust, easy-to-learn, easy-to-use software for wetlab scientists. Both require funding so they don’t become abandonware.
Tension: Research vs Surveillance vs Clinical (CLIA)
Public Health labs have three distinct mandates. They are the clinical labs of last resort within their states. They provide surveillance within their states. They work with research, often in partnership with local universities.
Clinical: It’s hard to get CLIA approval, and it must be done test by test. Every test requires complex quality control: sample intake standards, lab prep, sequencing, analysis and reporting. There is a direct tension with research because one cannot afford to re-validate an entire CLIA pipeline often (most labs re-validate once a year). There is a direct tension with surveillance, because the more rigorous and more time sensitive clinical tests crowd-out routine survellience. When urgent and important come into conflict, urgent (clinical) usually wins.
Tension: Policy paradoxes (Free-riding, Crowding-out)
Our Federal system of government distributes power between the Federal, State and Local governments. Under the theory of (top-down) Vertical Diffusion, the actions of the national government can influence — bias (generous interpretation) or distort (ungenerous interpretation)—the actions of State and Local government.
A sizable chunk of public health funding comes from the Federal government distributed to State and Local via grants and cooperative agreements. For example, a conservative estimate for FY2018 is that the CDC provided 12% ($7.5B of $64B) of State and Local public health funding [^2]. If the Federal government invests more in low capacity states, those states may respond by further defunding public health. Federal dollars thereby “crowdout” the money State and Local would have spent to protect their populations in the absence of Federal dollars.
If the Federal government requires matching funds to reduce “crowding-out” public health, states may reallocate funding between public health areas, still “crowding-out” but this time intra-public health.
Public health is a public good (non-rivalrous, non-excludable). Therefore the States may “”free ride” on the public health investments of other states — accuring some of the benefits without the cost of maintaining in-state capacity.
If the Federal government takes a “help those who help themselves”, the national government can reduce “free riding,” but further entrenches health inequities.
The Clinical Medicine Perspective
Observation: Uptake of NGS for clinical medicine has been slow.
There are opportunities around culture independent diagnostics and HIV resistance, but CLIA-approved test that rely on NGS are rare.
Observation: Segregation of clinical, public health lab, case and environmental data
These data are stored in different databases, are funded seperately, and usually reside in different government departments. This means that it is exceedingly difficult to link data between silos.
Further, the representation of case and environmental data is harder than lab data. If you have a well defined ontology, it’ll be too complex for Epi’s to use. If you rely on free-text notes, you won’t be able to computably link the datasets.
The Epidemiology Perspective
Observation: Cannot agree on a term, but Genomic Epidemiology is the most widely used
What do you call an Epidemiologist that knows how to use an interpret genetic data? I heard at least four different terms:
- Genomic Epidemiology
- Applied Genomic Epidemiology
- Biogenomic Epidemiology
- Human Genome Epidemiology (HuGE)
Unlike the more well established sub-fields of Bioinformatics or Pathogen Genomics, my observation is that the sub-field does not have a well codified body of knowledge.
The best codifications I have observed are…
- Pathogen: “An applied genomic epidemiological handbook” (actively updated)
- Human: “Human Genome Epidemiology (2nd ed.): Building the evidence for using genetic information to improve health and prevent disease” (archived / not actively updated)
Observation: Holy Wars
Every community has its holy wars. For software developers of a certain generation, it’s vim vs emacs. Within public health laboratories, it’s sequence everything vs selective sequencing (“responsible sequencing” for proponents).
Opportunity: Use NGS for things that take a long time using the traditional techniques
- Identification of non-culturable organisms
- Carriage studies
- Susceptibility testing
Appendices
Appendix: A brief history of CDC OAMD
2024: Relaunch AMD Academy with APHL and CSTE.
2023: We are here. 10-year anniversary of OAMD. CDC Reorganization [^3].
2022: Molecular Epidemiology Fellowship founded.
2021: SPHERES consortium founded, a distributed sequencing effort for COVID. Labs across US join Federal Initiative to Study Coronavirus Genome (New York Times). American Rescue Plan (ARP) funding. Fact Sheet: Biden Administration Announces $1.7 Billion Investment to Fight COVID-19 Variants.
2020: The COVID era. Public Health Grand Rounds Question: “Will you be applying NGS to COVID?”…Yes!
2019: NGS Quality Initiative founded. PulseNet transitioned from PFGE to NGS.
2018: All states have some NGS capacity.
2017: State Public Health Bioinformatics (StaPH-B) holds its first meeting at APHL Conference.
2016: 3/4 of states have NGS capacity. Bioinformatics Regional Resource (BRR) and AMD training leads established in 7 regions, funded by ELC grant.
2015: 1/2 of states have NGS capacity, driven by food bourne
2014: First AMD Day conference. CDC focused. AMD/APHL Bioinformatics Fellowship introduced.
2013: OAMD founded. $30M for FY14. Instruments co-located with academic labs.
2012: Budget memo on “Advancing infectious disease detection and response.” Request for $40M. Senate proposed $20M. Iterate through many names.
2011: Bioinformatics Blue Ribbon Panel
2010: Bioinformatics Working Group
2009: SciComp (on-prem high-performance computing cluster)
Appendix: Collateral
Centers of Disease Control and Prevention (CDC)
Congressional Research Service (CRS)
- crs-R47207: Centers for Disease Control and Prevention (CDC) Funding Overview (Updated March 28, 2023)
Government Accountability Office (GAO)
- gao-21–426: Genomic Sequencing of Infectious Pathogens
- gao-20–341: Antibiotic Resistance: Additional Federal Actions Needed to Better Determine Magnitude and Reduce Impact
Genomic Epidemiology Body of Knowledge
- Pathogen: “An applied genomic epidemiological handbook” (actively updated)
- Human: “Human Genome Epidemiology (2nd ed.): Building the evidence for using genetic information to improve health and prevent disease” (archived / not actively updated)
Footnotes
[^1] The New York City Department of Health and Mental Hygiene has six budget lines just for CDC ELC for FY2022.
4 ELC budget lines from the FY22 base appropriation: ELC COVID CARES, ELC COVID Expansion, ELC IPC Training Supplemental, ELC Supplement Enhanced Detect. Source: NYC Department of Health and Mental Hygiene, Fiscal 2022 Preliminary Plan, Appendices, A: Budget Actions in the November and the Preliminary Plans
2 additional ELC budget lines changes that occured durring FY22: CAT. ELC COVID, ELC COVID. Source: NYC Department of Health and Mental Hygiene, Health, FY23 Executive Budget, Appendix A: Budget Actions since Fiscal 2022 Adoption
[^2] The United States does not keep robust statistics on aggregate (federal, tribal, state, territorial, local) public health expendature, and the CDC budget is quite complicated. For example, the National Health Expenditure accounts overestimate public health funding. Therefore I provided a a back of the envelope computation for the FY2018. The $7.5B are grants from CDC to States; computed by summing from the CDC Grand Funding Profile. The $64B in aggregate state and local public health expendatures are the high-end estimate from Leider et al. ($35–64B).
[^3] OAMD was impacted by the CDC reorganization
- FROM: NCEZID > OAMD
- TO: NCEZID > Division of Infectious Disease Readiness and Innovation (DIDRI) > OAMD
- Where DIDIRI was itself reorged from Division of Preparedness and Emerging Infections (DPEI)
Source: https://www.cdc.gov/about/organization/cio-orgcharts/ncezid.html
Disclaimer: All views are my own, and do not represent those of my past, present or future employers.
