Identifying Contaminants Before Filtering: E coli?
Hi! I am trainee who is new to Qiime2, and I'm trying to decide whether or not to filter out E coli before proceeding with my core-metrics analyses. I need a little more information than the filtering tutorials provide to make my decision.
For context, my samples are from 2 different tumor types retrieved from various locations throughout the human body, and my team sequenced reads from 16S V3-V4. We are hoping to determine the enriched species present in each tumor type, and secondarily, we want to see if those enriched species vary by location. In my unfiltered taxa bar plot, I can see that almost every sample contains a sizable chunk of E coli reads. This includes samples from locations where E coli is expected (ie GI samples) and locations where it is less expected (ie sinonasal samples).
Here's a screenshot of my unfiltered bar plot with the relative frequency of E coli in each sample highlighted:
I guess I'm a little unsure on whether to --p-exclude g__Escherichia because 1) E coli is a common contaminant, 2) I may obscure the significance of less-abundant taxa if I keep it in, and 3) it's present in such large chunks in my samples that it will definitely affect my diversity analyses. However, I'm also wondering if E coli could be a legitimately enriched species in these tumor types because the species has been linked to other cancers (ex - pks+ E coli in colon cancer). I asked my PI, and he said, "It's a gray area."
What are your thoughts? If anyone could suggest papers that provide guidelines for identifying & filtering contaminants or offer a reviewer-like insight on how this would affect my downstream analyses, I would greatly appreciate it.
I usually work with fungi rather than bacteria, but here are my two cents until someone with more experience weighs in
Yes, I think that if you just remove E. coli globally, you would be assuming that the E. coli in the gut is just as fake as the one in the nose/sinuses, and that would definitely bias your downstream diversity analyses.
That being said, I'm guessing that since you sequenced tumors, you were expecting low bacterial DNA, so it is safe to suspect that the E. coli in those low-biomass samples might indeed be a contaminant. If you sequenced some negative controls (e.g. blanks),you can check them for E. coli presence. If not, maybe you can plot your starting DNA concentrations against E. coli relative abundance. If you see a strong negative correlation, it would be sensible to guess that it is contamination.
Also, if you sequenced e.g. your gut samples on the same lane as your sinonasal samples, you might be experiencing index hopping. To my knowledge, this is not really likely to happen, but maybe you want to consider it.
The q2-quality-control plugin has some great methods to deal with contamination, such as decontam-identify. Maybe it is worth giving it a try before your next step?
I appreciate you for taking the time to reply. I will take a look at those plugins - thank you for providing those easy links. In the meantime, I will probably also take a look at the differences in diversity observed with E coli included vs excluded, although I'm not sure which result I would trust.
As a follow-up, if I don't have access to the concentrations of each sample, is it appropriate to use the total number of reads from the sample as a proxy?
Unfortunately, no. The final number of reads you obtain depends on the sequencing depth. Sequencing facilities typically try to get a similar amount of reads for all samples, so the final read count is decoupled from your initial DNA concentration.
Sounds like a good plan. It will at least show you how much of your beta diversity variance is being driven by this single taxon. Just be sure you don't make a decision based on the results you "like" the most for your story, but based on the methodology that makes more sense for you
Good luck! If you need anything else, please let us know
forum!