Thread by @NathanGrubaugh

📢 PSA for GISAID submitters/users

All of our B.1.621 + B.1.621.1 (important emerging SARS-CoV-2 variant) submissions are being rejected by GISIAD. This has some important implications that data producers + public health agencies should be aware of.

Technical 🧵(1/12)

We usually have a few sequences/week that get rejected for QC reasons (eg indel in a string of As or Ts). It takes a few days for these to be fixed and reposted. Sometimes longer depending on our bandwidth. These are usually a random distribution so not a problem if delayed(2/12)

This past week we had 40+ sequences get rejected, and almost all of them were B.1.621/B.1.621.1. @JosephFauver found that all of these sequences have a 4 nt deletion in ORF3A that results in a premature stop codon about ~50 nt upstream. GISAID sees these and kicks them back(3/12)

Why is this important?

Some groups may not have the bandwidth to investigate and approve rejected sequences and thus they don't get uploaded to GISAID. Other times it means that these data are delayed by several days. (4/12)

In both cases they bias the data.

- Inflates the proportions of all other lineages, including Delta, by changing the denominator

- Deflates the proportions of B.1.621 + B.1.621.1

Depending on the location, the overall impact could be quite large. (5/12)

We do the majority of the sequencing for Connecticut, Puerto Rico, US Virgin Islands, and the Dominican Republic. We are finding that B.1.621 + B.1.621.1 recently emerged in these regions and at times can make up ~10% of the sequenced cases. (6/12)

But yet if you go to outbreak[.]info - which pulls its data from GISAID - you wont find any B.1.621 sequences from Connecticut because we are still working to get GISAID to accept them. And if you look at the entire US, it shows that B.1.621 is <1%. (7/12)

I suspect that B.1.621 + B.1.621.1 are actually much higher than 1% in the US and are significantly under-reported across the world.

Outbreak[.]info classifies B.1.621 as a 'Variant of Interest' as it has E484K, N501Y, and P681H found in other important variants. (8/12)

Beyond missing the emergence of a potentially important variant, the rejection of B.1.621 submissions can also inflate the frequency of other variants.

For example, last week we reported that Delta in CT was <70% while the @CTDPH report said ~80%. (9/12)

The difference between our frequencies is that we use data directly produced by our lab and reported to us by our partners and the @CTDPH only uses data they find on GISAID (which includes our data). Since B.1.621 is being rejected by GISAID, their denominator is off. (10/12)

The @CTDPH is aware of this issue and we are working with GISAID to correct it. I'm telling y'all because this is likely an issue in other places too.

**Please be aware of potential biases of using only data pulls from GISAID for surveillance** (11/12)

Recap:
- GISAID is rejecting B.1.621 sequences due to a real 4 nt deletion in ORF3A
- Please try to confirm these as accurate ASAP so limit the data bias that these rejections may cause.

Hopefully GISAID will fix this rejection issue soon. (12/12)