Outliers detection and treatment
Otávio Lovison
Lucas Czech
Lucas
Otávio Lovison
--
You received this message because you are subscribed to the Google Groups "Phylogenetic Placement" group.
To unsubscribe from this group and stop receiving emails from it, send an email to phylogenetic-plac...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/phylogenetic-placement/34c05cae-40c7-4baf-91bf-65299f98dc82n%40googlegroups.com.
Lucas Czech
Hi Otávio,
for sample outliers like that, I think my PhD thesis might help: https://publikationen.bibliothek.kit.edu/1000105237
It contains a more thorough examination of the methods of gappa (more than the corresponding papers). See in particular Figure 4.5(a). There is a black arrow pointing at one particular edge - which comes from an outlier sample.
From the text:
Further examples of variants of Edge Dispersion on the BV dataset are shown in
Figure 4.5. In Figure 4.5(a), which is linearily scaled, it is striking that one outlier
edge, marked with an arrow, is dominating the values, and thereby hiding the values
on less variable edges. This outlier occurs for the species Prevotella bivia in one of the
220 samples, where 2781 out of 2782 sequences in the sample have some placement
mass on that branch. Upon close examination, this outlier can also be seen in
Figure 1D of Srinivasan et al. (2012) [339], but is less apparent there. Thus, our
novel visualization can help to detect such outlier samples.
Hope that gives you some idea or direction ;-)
Other than that, this is up to you now - you might be able to work out a novel method there!
As for chimeras: No idea, that's unsolved as far as I am aware. If you have an idea, let me know!
For the tax assignment: That depends on what you want to achieve there. Just assignment of your queries to the most likely taxonomic labels? Or more?
Cheers
Lucas