Closed ktmeaton closed 6 months ago
On a related note, with the default --depthcutoff 0
, is the lineage abundace getting evenly split between all possible lineage candidates?
BA.1-like
= 99.9% -> BA.1.1
= 49.9% + BA.1.4
= 49.9% ?
In this particular sample, there is 0 coverage at the lineage disciminating sites: G22599A
and G23351A
. Which I'm interpreting as this sample could be either BA.1.1
or BA.1.4
. Does splitting ir risk underestimating the abundance?
(Sorry if these questions are already documented somewhere else!)
Hey @ktmeaton!
Thanks for bringing this up. At present the summarized output doesn't properly use the collapsed groupings generated via use of the --depthcutoff
parameter, but this is something we're hoping to incorporate soon.
Relatedly- the groupings present issues for plotting, as collapsed groups can differ across samples. We're planning to leverage the group MRCAs to make this work relatively soon.
And yes, if Freyja is unable to differentiate between lineages given the available coverage, it will return equal fractions of the possible lineages back to you. The correct thing to do in this case IMO is to instead collapse into a BA.1-like lineage, that includes BA.1.1 and BA.1.4.
Josh
Thanks so much for the response! I was thinking that collapsing to BA.1-like was the preferred approach, but just wanted to check.
Regarding MRCAs, do you have any wisdom/experience to pass along about getting MRCAs between recombinants? Especially for those crazy nested ones like XDM (XDA + GW.5)? I've attempted to write some functions that follow all possible ancestors, but I'm not sure if that's the right approach, and it's dizzying to test. And then breaking ties on top of that gets complicated (ex. XE and XG would have multiple MRCAs? BA.1 and BA.2?). Just curious what you've tried for freyja!
No prob! Yeah, it gets pretty tricky when dealing when we get into recombinants. For the nested ones, generally you can break things down the nested ones by pulling the parents of recombinant, and then looking for overall MRCA. Of course if those parents are pretty distinct (like a Delta/Omicron recombinant, the MRCA heads towards the root pretty fast).
Since we're still in the development phase of that functionality - right now a lot of these cases with recombinants will result in a "Misc" group name, but hopefully more to follow up on that soon!
Hi Freyja Team,
Is it intentional that when using the
--depthcutoff
parameter, the-like
lineages are all grouped within theOther
summarized? For example, if I disable the cutoff (--depthcutoff 0)
, then I can get an ambiguous mix ofBA.1.1
andBA.1.4
for a total of 99.9%Omicron
. But if I enable the cutoff (--depthcutoff 1
), regardless of value, I getBA.1-like
that is 99.9%Other
.Ideally, I think I'd want to see the
-like
lineages counted towards their known VOC (ex. Omicron). But is it intentional that they're grouped under Other instead?v1.4.9
02_16_2024-00-48
SRX19564628
through ViralRecon and downsampled it for testing. SRX19564628.depth.txt SRX19564628.variants.txtCommands
Place sample data in
data/
.Download the freyja database:
Run freyja demix: