1 of 32

Supporting a nationwide Covid-19 wastewater monitoring program

Claire Duvallet, PhD

Biobot Analytics

claire@biobot.io

2 of 32

Biobot Analytics

We are building early warning health analytics from data available in our sewers.

Mariana Matus, PhD

CEO and Cofounder

mariana@biobot.io

Newsha Ghaeli

President and Cofounder

newsha@biobot.io

3 of 32

We started by addressing the opioid epidemic

3

4 of 32

As a public health company, we responded to Covid-19

4

5 of 32

Academic collaboration to develop & validate method to detect SARS-CoV-2 in wastewater in March 2020

5

Wu, Zhang, Xiao et al. “SARS-CoV-2 Titers in Wastewater Are Higher than Expected from Clinically Confirmed Cases.” mSystems. 2020. doi: 10.1128/mSystems.00614-20

Validation results: SARS-CoV-2 wastewater titers, Boston March 2020

6 of 32

Biobot’s first nationwide deployment, March-June 2020

Wu, Xiao, et al. “Wastewater surveillance of SARS-CoV-2 across 40 U.S. states.” Water Research. 2021. doi:10.1016/j.watres.2021.117400

6

7 of 32

Broad adoption of WBE across academic and public health communities

So many papers!

Case studies: communities and universities
Methods development, comparison, optimization
Validation: wastewater correlates with cases!
Interpretation: leading indicator, WC ratio, incidence vs. prevalence
Variant detection and sequencing

So much adoption!

Local municipalities
State-level programs funded by CDC, EPA
National Wastewater Surveillance System launched (!!)
HHS Covid-19 wastewater monitoring program

7

PubMed results for wastewater-based epidemiology

8 of 32

HHS Covid-19 Wastewater Monitoring Program, June 2021

In collaboration with Biobot Analytics

9 of 32

HHS Covid-19 Wastewater Monitoring Program

The HHS Covid-19 Wastewater Monitoring Program was a partnership between the US Department of Health and Human Services (HHS), the Centers for Disease Control and Prevention (CDC), the National Institutes of Health (NIH), and Biobot Analytics
It goal was to assess the amount of COVID-19 in our communities by testing wastewater samples from across the country.
The program also generated sequencing data on viral RNA from samples to support R&D on detecting and monitoring variants.
This program was part of federal efforts to expand a national wastewater monitoring system for COVID-19. The program was designed to cover over 100 million people in 50+ states, US territories, and Native American communities.

9

10 of 32

How it works

10

2

Sample collection and logging

Lab analysis

Computational biology and data science

Report

Participant takes a 24-hour composite influent sample, logs the data, and ships it back using the provided FedEx label. Sample arrives at our lab the next day.

Molecular biology analyses, applying Biobot’s protocol in high-throughput.

Data analysis and visualization is then packaged as a report, which is delivered back to participants.

We process and interpret your data with the latest models derived from our entire dataset.

3

4

5

1

Biobot sends participants kits

Our team sends you all of the sampling kits you need to participate.

11 of 32

Biobot onboarded 300 WWTPs in 3 weeks

Pre-existing relationships

Professional communities

Direct connections via HHS/CDC

Outreach

11

12 of 32

Biobot onboarded 300 WWTPs in 3 weeks

Pre-existing relationships

Professional communities

Direct connections via HHS/CDC

Outreach

Communication

Onboarding & data interpretation webinars

Responsive customer success team

PDF Reports delivered to participants

12

13 of 32

Biobot onboarded 300 WWTPs in 3 weeks

Pre-existing relationships

Professional communities

Direct connections via HHS/CDC

Outreach

Communication

Onboarding & data interpretation webinars

Responsive customer success team

PDF Reports delivered to participants

Logistics

Pre-assigned sampling days

Shipped kits in bulk

Sampling site metadata & sampling log forms

13

14 of 32

Sampling location metadata form

Online metadata form: https://www.app.biobot.io/sampling-location-metadata

14

As many of you know, collecting metadata is actually a really challenging aspect of doing WBE at scale.

We gave participants a unique ID when they signed up to participate, and then pointed them to a form to collect sampling location metadata.

They could register new sampling locations here, and provide all the metadata required for data interpretation and submission to NWSS.

This form was also important to us, since we were being evaluated based on population coverage, which was provided here.

We worked hard to build these forms to be user friendly and intuitive. One lesson learned was the balance between comprehensive data collection and lower barriers to filling out the data.

We didn’t gate sending samples on getting this info, which also required follow-up if some participants didn’t fill it out. At the end, though, we got metadata from every participant.

15 of 32

Sampling log

Online sampling log form: https://www.app.biobot.io/sample

Participant enters Kit ID, confirms organization, and fills out sampling log information.

15

16 of 32

SARS-CoV-2 quantified in-house at Biobot’s lab

Pasteurize wastewater sample�one hour at 60 °C
Concentrate RNA �Amicon centrifugal ultrafiltration
Extract RNA�RNeasy columns or cassettes
Quantify SARS-CoV-2�Multiplex N1/N2 RT-qPCR
Quantify fecal marker & controls�Multiplex PMMoV/BRSV/IAC RT-qPCR

16

Duvallet, Wu, Imakaev, McElroy, et al. “Nationwide trends in COVID-19 cases and SARS-CoV-2 wastewater concentrations in the United States.” medRxiv. 2021. doi: 10.1101/2021.09.08.21263283

17 of 32

All samples pass robust laboratory & data QA/QC

Negative (no template) controls and extraction blanks

Positive controls for N1/N2 and PMMoV

PMMoV used to flag low recovery

Laboratory controls

Sample controls

PMMoV as endogenous sample control

qPCR curves manually reviewed for inhibition

Spike-in recovery and exogenous inhibition controls

Data review

First sample for a new location

Large increase or decrease since last sample

Sampling log suggests shipping delay

17

18 of 32

Contract also included a sequencing component

Sequencing data generation

Sequence one eligible sample per location per week

Only sequence if Ct < 37

Sequencing data generation outsourced

Ginkgo Bioworks
RNA extracted at Biobot
ARTIC v3 library prep + NextSeq550

Data uploaded to NCBI

BioProject PRJNA746354

18

As I mentioned, this contract also included a sequencing component.

We sent one sample per week per location, screening out samples with too low of SARS-CoV-2 for the sequencing to be successful.

We partnered with an outsourced lab to generate the data, and uploaded all of our sequencing data to NCBI.

We also did some variant analysis, again using pretty standard methods that you’ve heard about here.

First step was to identify variant-specific mutations by looking at the clinical data in GISAID, and then looking for those mutations in our WW samples.

We aligned the fastq data to a reference SARS-CoV-2 genome and considered samples a sequencing success if at least 20% of the genome mapped at >10x coverage. To quantify the Alpha and Delta variants in these samples, we first identified a set of variant-specific mutations. We downloaded all US sequences submitted to GISAID before 1 Sept. 2021, and identified mutations which were present in >80% of sequences of the variant of interest and absent in >90% of sequences associated with other variants. We then counted the percentage of reads with the Alpha- or Delta-associated mutation at each of the characteristic mutation locations. Finally, we averaged this percentage across characteristic mutation locations to get a sample-wise relative abundance of Alpha and Delta.

19 of 32

Contract also included a sequencing component

Sequencing data generation

Variant analysis

Sequence one eligible sample per location per week

Only sequence if Ct < 37

Sequencing data generation outsourced

Ginkgo Bioworks
RNA extracted at Biobot
ARTIC v3 library prep + NextSeq550

Data uploaded to NCBI

BioProject PRJNA746354

Identify variant-specific characteristic mutations

Using clinical US data from GISAID
Present in majority of variant sequences & absent in majority of non-variant sequences

Align WW data to reference genome
Estimate variant percentages per sample

Count percentage of reads at each characteristic mutation
Average across all mutations

19

As I mentioned, this contract also included a sequencing component.

We sent one sample per week per location, screening out samples with too low of SARS-CoV-2 for the sequencing to be successful.

We partnered with an outsourced lab to generate the data, and uploaded all of our sequencing data to NCBI.

We also did some variant analysis, again using pretty standard methods that you’ve heard about here.

First step was to identify variant-specific mutations by looking at the clinical data in GISAID, and then looking for those mutations in our WW samples.

We aligned the fastq data to a reference SARS-CoV-2 genome and considered samples a sequencing success if at least 20% of the genome mapped at >10x coverage. To quantify the Alpha and Delta variants in these samples, we first identified a set of variant-specific mutations. We downloaded all US sequences submitted to GISAID before 1 Sept. 2021, and identified mutations which were present in >80% of sequences of the variant of interest and absent in >90% of sequences associated with other variants. We then counted the percentage of reads with the Alpha- or Delta-associated mutation at each of the characteristic mutation locations. Finally, we averaged this percentage across characteristic mutation locations to get a sample-wise relative abundance of Alpha and Delta.

20 of 32

Overall, we achieved a successful program!

>6,000 samples collected

>2,000 samples sequenced

Participants from across the US:

All 50 states and DC
2 US territories
9 Tribal Indian territories

Sampling covered 90 million people

More than 27% of all Americans

Participants from different community sizes:

114 plants covering <50,000 people
162 plants covering 50,000–500,000
54 plants covering >500,000

Number of HHS Program samples collected by state

20

21 of 32

Different ways to define coverage

21

22 of 32

Wastewater monitoring is equitable

Wastewater monitoring better represented the age structure and racial demographics of the US population compared to the vaccinated population.

22

https://biobotanalytics.medium.com/wastewater-for-equitable-covid-19-monitoring-e3b947d91e0a

We also wanted to look at how representative wastewater monitoring was throughout this program.

We compared the demographics of populations covered by WWTPs participating in this program, vs demographics of vaccinated and total US population.

Wastewater monitoring better represented the age structure and racial demographics of the US population compared to the vaccinated population

Age:

Younger people are underrepresented among the vaccinated population, while older people are overrepresented.
This sort of skew in age groups makes sense for vaccination, since older adults are at highest risk for hospitalization and death.
But all age groups can spread the virus, so it is important for disease surveillance to have an even representation across age groups, which wastewater monitoring can provide.

Race:

Looking at racial demographics, we know that Black and Hispanic communities are disproportionately affected by Covid-19.
These demographics were actually slightly over-represented in our dataset relative to the entire US population, and especially relative to vaccinated populations.
Importantly, Native American populations were also an important part of the HHS project - explicitly called out as communities that needed to be included.
We ended up with nine tribal communities spread over six different states participated in the project.

23 of 32

Cases correlate with wastewater well (!!)

Shown is the average of two samples per week
Wastewater correlates with cases over 2 orders of magnitude

23

24 of 32

Wastewater and cases reflect summer surge

24

Wastewater

Clinical

cases

25 of 32

Delta sweep is also seen in wastewater

25

26 of 32

Delta sweep precedes SARS-CoV-2 increase in wastewater

26

27 of 32

Variants in wastewater reflect clinical dynamics

27

Clinical

Wastewater

28 of 32

Lessons learned reinforce the groundwork for continued nationwide wastewater programs.

Some key lessons learned, or re-enforced:

Data sharing: Participants must have immediate access to –and control over– their own data. Sharing data with local public health officials, elected officials, and researchers should be easy.

28

Beyond the results though, we’ve learned some key lessons about implementing wastewater surveillance at scale.

Over and over, a key theme that emerged from participants was that sharing data back with them promptly was so important for their continued buy-in. Also making it easy for them to share data with their elected officials and public health officials was super important - many of our sampling partners were grateful that we provided high-level data interpretations on our reports, so that they could simply forward them to less technical stakeholders in their communities.

We’ve also learned that data quality goes far beyond the lab: collecting and QCing metadata is in some ways just as important as QCing the results that come from the lab. Simply collecting the metadata for all 300+ locations was a huge effort, and that’s why it’ll be important to have standardized interfaces and collection systems so that utilities don’t have to keep uploading their metadata over and over.

And finally, we know that just handing over lab results isn’t enough for wastewater to get acted on. The results have to be paired with training and data interpretation support, so that we can ensure that when it gets into the hands of folks who are empowered to act on it, the know how to.

29 of 32

Lessons learned reinforce the groundwork for continued nationwide wastewater programs.

Some key lessons learned, or re-enforced:

Data sharing: Participants must have immediate access to –and control over– their own data. Sharing data with local public health officials, elected officials, and researchers should be easy.
Data quality: Collecting and reporting metadata isn’t always easy. A nationwide program must have a simple, consistent metadata upload interface.

29

Beyond the results though, we’ve learned some key lessons about implementing wastewater surveillance at scale.

Over and over, a key theme that emerged from participants was that sharing data back with them promptly was so important for their continued buy-in. Also making it easy for them to share data with their elected officials and public health officials was super important - many of our sampling partners were grateful that we provided high-level data interpretations on our reports, so that they could simply forward them to less technical stakeholders in their communities.

We’ve also learned that data quality goes far beyond the lab: collecting and QCing metadata is in some ways just as important as QCing the results that come from the lab. Simply collecting the metadata for all 300+ locations was a huge effort, and that’s why it’ll be important to have standardized interfaces and collection systems so that utilities don’t have to keep uploading their metadata over and over.

And finally, we know that just handing over lab results isn’t enough for wastewater to get acted on. The results have to be paired with training and data interpretation support, so that we can ensure that when it gets into the hands of folks who are empowered to act on it, the know how to.

30 of 32

Lessons learned reinforce the groundwork for continued nationwide wastewater programs.

Some key lessons learned, or re-enforced:

Data sharing: Participants must have immediate access to –and control over– their own data. Sharing data with local public health officials, elected officials, and researchers should be easy.
Data quality: Collecting and reporting metadata isn’t always easy. A nationwide program must have a simple, consistent metadata upload interface.
Data interpretation: Just handing over the raw results isn’t enough. A nationwide program must include training and interpretative support, to enable action.

30

Beyond the results though, we’ve learned some key lessons about implementing wastewater surveillance at scale.

Over and over, a key theme that emerged from participants was that sharing data back with them promptly was so important for their continued buy-in. Also making it easy for them to share data with their elected officials and public health officials was super important - many of our sampling partners were grateful that we provided high-level data interpretations on our reports, so that they could simply forward them to less technical stakeholders in their communities.

We’ve also learned that data quality goes far beyond the lab: collecting and QCing metadata is in some ways just as important as QCing the results that come from the lab. Simply collecting the metadata for all 300+ locations was a huge effort, and that’s why it’ll be important to have standardized interfaces and collection systems so that utilities don’t have to keep uploading their metadata over and over.

And finally, we know that just handing over lab results isn’t enough for wastewater to get acted on. The results have to be paired with training and data interpretation support, so that we can ensure that when it gets into the hands of folks who are empowered to act on it, the know how to.

31 of 32

The work continues…

https://biobot.io/data/

31

32 of 32

Claire Duvallet

Founding Technical Lead, Data Science

claire@biobot.io

Max Imakaev

Staff Data Scientist

max@biobot.io

Katherine Stansifer

Group Lead, Computational Biology

katherine@biobot.io

Scott Olesen

Group Lead, Epidemiology

olesen@biobot.io

Thank you

Biobot Analytics

www.biobot.io

hello@biobot.io