Thread by @SolidEvidence

GISAID vs SRA/WW
I thought I would do a little comparison to see how wastewater sequencing data compares with patient sequencing data in evaluating viral trends.
1/
cdc.gov/nwss/index.html

For WW I took all of the samples from our most recent SRA download that were collected in the last month (~500 samples). This wasn’t normalized.

For the patient side I used Cov-Spectrum data (because it's public) from the last month (8,302 sequences).
2/
cov-spectrum.org/explore/World/…

There are about 50k patient samples collected for sequencing each month, but there is always a delay before they are all sequenced and uploaded.

In this regard, the WW data is much faster.
3/

Example: Verily (who has CDC/NWSS contract) already has 145 samples available from the last 2 weeks.

By contrast, there are only 13 US patient sequences available that were collected in the last 2 weeks.

Plus, each WW sample represents 10s to 100s of thousands of people.
4/

From the SRA/WW samples there were 80+ changes in Spike that were essentially consensus. These perfectly matched the changes in JN.1.

Also near consensus were F456L (99%) and Q493E (90%).

From the patient sequences F456L was 96% and Q493E was 89%.

Not a bad match.
5/

Next on the WW list was S31- (68%), T22N (20%), and F59S (18.5%).
From the patient sequences S31- was 61%, T22N (30%) and F59S (27%).

T22N/F59S is mostly XEC. The small disconnect is probably because more of the WW data is mostly from the US where XEC isn't as prevalent.
6/

US sequences from the last month were: 31- (69%) , S22N (24%), and F59S (21%).
7/

The 31- includes KP.3.1.1*, but also a bunch of other lineages.

To estimate KP.3.1.1 prevalence I looked outside of spike. 13,121T was 58%. (12,616T had low coverage).

In patients 12,616T and 13,121T were both 56%.

Good agreement, KP.3.1.1 at ~56-58%.
8/

Next from WW/SRA were R346T (17.6%), T572I (8.1%), Q183H (6.9%) and H146Q (6.4%).
Numbers for patients were 10%, 10%, 3%, and 1%, respectively.
Not sure why there is a bit more of a disconnect with these, but I’m guessing it is because the patient data is behind. We’ll see.
10/

Here are the rest of the WW changes at 2% or higher. Despite my prognosticating, F456V (MV.1, not listed) is still only at 0.2%.
11/

It's worth noting that in addition to being fast, wastewater sequencing is pretty cheap compared to patient sequencing.
12/

All in all, I'd say the WW data does quite well at getting a fast, cheap and accurate overview.

The future of wastewater surveillance is not certain, but I hope future administrations recognize that it is an efficient and cost-effective means of monitoring pathogens.
13/13