When we only look at the viruses specified above, we find that plasma contains 18 out of the 21 viruses, whereas whole blood contains 17 out of the 21 viruses. Specifically, plasma uniquely contains HBV, whereas whole blood uniquely contains HTLV-1 and HTLV-2.
I would also give the total number of matches for some of these viruses. If there are frequently only one match, than that could be spurious (BLASTing them could help in making the matching more confident)
(this is in accordance to the NAO’s interest to detecting “stealth” pandemics),
cut
Figure 4: Relative abundance of selected human-infecting viruses in each dataset. Top: without controlling for human reads. Bottom: controlling for human reads.
I would pick either the controlled for human RA or the not controlled for human RA numbers here.
I would add alpha here (i.e., make the symbols slightly transaprent).
I find it quite hard to quickly match the different symbols with the different studies.
I would add more y tick labels
The horizontal grid lines are slightly confusing, as they do not help in matching the symbols to the labels on the left.
I would cut down on the number of viruses covered. E.g., you could have a cut off of ten samples minimum for each virus.
Figure 3: Family composition of human-infecting viral reads
This is cool. I would add information on how many HV reads make up each of these bars.
Table 5: Relative abundance of human-infecting viruses in each sample type
What kind of average are you using here?
1.82e-04
These numbers are too small given the plot of human reads above.
reads
total reads?
RA (control for human reads)
Use the same metric for RA everywhere (above you use percentage)
sample_type
would add a column for enriched vs non-enriched
We can try to get a better understanding of the performance of whole blood by controlling for human reads (Table 3). The number of viral reads becomes much higher when we control for human reads, making it comparable to the plasma datasets.
simplify wording
particularly if looking at the extracellular content.
say why
Even comparing viral enrichment methods, we can see that Cebria-Mendoza has a much higher fraction of viral reads compared to Thijssen.
Easier to double-check this by having enriched studyies marked accordingly
Choice of viral enrichment tends to have an impact on kingdom composition, as we can see that Cebria-Mendoza and Thijssen, both of which did viral enrichment, have significanlty lower human read fractions than Mengyi, despite all datasets coming from plasma.
Improve wording
4.1 Kingdom composition
Would also indicate which of these bars represent a study that is using panel-amplification.
modifyed
modified
This approach provides comprehensive genetic information but precludes separate analysis of DNA and RNA from individual samples.
What kind of sequencing did they do (single-end or paired-end, what bp length)
This analysis will help us understand the last piece missing from the puzzle
For the final blog post, I'd change this wording. Don't know what puzzle you are talking about here.
we identfied four sampling strategies, each coming from either whole blood or plasma.
Would reiterate what those are.
About
What was the original goal of the study