Thread by @LauringLab

Seeing lots of analyses scraping GISAID data and looking at variant sequences / total sequences. I've heard chatter in my corner of the world that "Wow, Michigan really has a variant problem." Do we? Short thread about putting the epidemiology back in genomic epidemiology 1/x

In Michigan, who uploads sequence data? Vast majority comes from our awesome State Lab and then my lab. Since 1/1, my lab has uploaded 1146 to GISAID. Don't know State Lab numbers but at least 2x that. There are 600+ B.1.1.7 from Michigan, according to CDC 2/x

Why is that important? Well you need to know how sequencing is done. My lab is sequencing broadly in our area - we do every positive that we can get our hands on. We've been getting sequences uploaded on > 25% of cases since September. We've done better since January. 3/x

So that is a really good sample, albeit from a smallish region. The State Lab sequences broadly across the State, but will also do a lot of sequencing for specific contact traces or outbreaks. This is important for the data. 4/x

If you sequence more, you will find more. So total variant cases can be misleading if you don't control for number of cases or number of cases sequenced. But how cases are selected for sequencing can also lead to bias. See this note on outbreak.info. 5/x

How does the splay out in Michigan. Well. Look at this, also from outbreak.info. See that huge spike in percent of cases with variant? What's up with that? 6/x

State lab sequenced a major outbreak of B.1.1.7 in a prison. Don't know final numbers, but at least 90 cases from the very start of the outbreak. This is major ascertainment issue. State sequences an outbreak is not representative sample. 7/x
freep.com/story/news/loc…

How do things look with a broad sample from a small area (Washtenaw County, 367,000 residents with a major University)? We've had 6.5% of our sequences as B.1.1.7, most tied to an outbreak in our campus community. 8/x

The last two sequencing runs (weekly) of 192 samples, we've had approximately <10% B.1.1.7. 9/x

Hope you've found this interesting. It's important to understand the data you're using in order to draw conclusions. Especially important if data are being used for policy decisions. One way to get a better sense of the data is to actually ask the people who generate it. 10/x

To answer my question, "Do we have a variant problem?" Maybe. Certainly would like there to be fewer variants around my neck of the woods. But, not sure it's worse in Michigan than elsewhere around the country. 11/x

Adding to this thread. Folks pointing to helix data. I’ve learned that this is non overlapping with data from state public health labs (which in MI is substantial). Also can’t figure out denominator on helix site (percent of how many?) or how they sample States. Anyone know?