I have a very special problem. I've got data only about the abundance of specific phyla in a patient cohort. That means that I know how many of each Phyla a patient has, like:
Patient 1: 10 of phyla 1, 21 of phyla 2, 300 of phyla 3, ... etc.
Patient 2: 15 of pyhla 1, 26 of phyla 2, 365 of phyla 3, .... etc.
...
However I want to go on with microbiome analysis but therefore I need OTUs. However, OTUs are used to classify bacteria at the genus level, but now I have the data on phyla level and I have no 16S sequencing data. How can I create an OTU only by having the phyla abundance? Is is possible?
Do I really need OTUs? Can I analyse alpha-/beta-diversity without OTUs?
How can I create an OTU only by having the phyla abundance? Is is possible?
You can't get an OTU table from a taxonomic table with abundance at phylum level.
Do I really need OTUs? Can I analyze alpha-/beta-diversity without OTUs?
An OTU is just a taxonomic unit with the lowest taxonomic resolution as possible. The idea is to get an approximation for species. Doing alpha/beta-diversity without species level abundance data or its approximation, OTUs (or other taxonomic unit such as ASVs - Amplicon Sequence Variants) is not really meaningful or the point of the analysis. This is because you usually do alpha-/beta-diversity to study a microbial community. When you think well, a microbial community is comprised by several species with different number of individuals.
Although there are perhaps a few analyses that you might do that give you an idea of alpha and beta-diversity of the community. Regarding alpha-diversity, you can discuss how many different phyla were found in each sample. This gives you an idea of how much within alpha-diversity you have in one sample. Keep just in mind that if you look at a different taxonomic level or even at OTU level, the diversity can be different than at phylum level. Regarding beta-diversity, you can do an hierarchical heatmap, i.e., an heatmap with an hierarchical clustering, clustering the samples based on phyla distribution similarity across the different samples. This allow you to have an idea of between sample diversity. Although I would like to warn you for the same issue that I meantion previously: if you do the exact same analysis at a different taxonomic level or OTU level, you might get a different result and, that's why this kind of analyses are usually done at lower taxonomic resolution as possible - OTU level.
Thanks Antonio! That was an excellent explanation! Especially the idea of doing an hierarchical heatmap is really good. I will try to focus on that. That's really interesting, that alpha- and beta-diversity and even OTUs only refers to species level. But nevertheless, is it able to construct phylogenetic trees only having data on the phylum level? Or calculate any kind of "metric distances" based on the taxonomic composition/diversity?
But thank you very much for your really detailed answer. This helped me a lot!
is it able to construct phylogenetic trees only having data on the phylum level?
Only if you have the original 16S rRNA amplicon gene sequences, which I think you don't have. Otherwise you could process the data set in order to get OTU table.
Or calculate any kind of "metric distances" based on the taxonomic composition/diversity?
This is somehow done in the suggested hierarchical heatmap. During the hierarchical clustering step there is some (dis)similarity metric that you need to use to cluster the data, e.g., Euclidean distance (there are many), which is based on the taxonomic composition/diversity, which in your case will be based on phyla level composition.
Thanks again Antonio for your quick and very helpful answer! I really need to study more about clustering but I think I understand the procedure thanks to your explanations!
Otherwise you could process the data set in order to get OTU
This sounds really interesting. So I can really process my data to get OTU? You are correct, I don't have 16S rRNA amplicon gene sequences, so I thought this was impossible. How does it work to get OTU out of my really slim dataset?
Sorry, perhaps I was not clear enough. No, you can't get OTU level assignments from a table summarised at phylum level.
What I was trying to say was, to do a phylogenetic tree you need to have the 16S rRNA gene sequences, and if you have them, you could process them to get an OTU table. Since you don't have 16S rRNA gene amplicon sequences for this data set, you can't do a phylogenetic tree neither obtain a table at OTU level.
There are several ways to obtain OTU tables, which rely on clustering the sequences at some identity threshold level. If you don't have the sequences, you can't cluster them, and, therefore you can't obtain a OTU abundance table.
Thanks Antonio! That was an excellent explanation! Especially the idea of doing an hierarchical heatmap is really good. I will try to focus on that. That's really interesting, that alpha- and beta-diversity and even OTUs only refers to species level. But nevertheless, is it able to construct phylogenetic trees only having data on the phylum level? Or calculate any kind of "metric distances" based on the taxonomic composition/diversity?
But thank you very much for your really detailed answer. This helped me a lot!
You're welcome!
Only if you have the original 16S rRNA amplicon gene sequences, which I think you don't have. Otherwise you could process the data set in order to get OTU table.
This is somehow done in the suggested hierarchical heatmap. During the hierarchical clustering step there is some (dis)similarity metric that you need to use to cluster the data, e.g., Euclidean distance (there are many), which is based on the taxonomic composition/diversity, which in your case will be based on phyla level composition.
António
Thanks again Antonio for your quick and very helpful answer! I really need to study more about clustering but I think I understand the procedure thanks to your explanations!
This sounds really interesting. So I can really process my data to get OTU? You are correct, I don't have 16S rRNA amplicon gene sequences, so I thought this was impossible. How does it work to get OTU out of my really slim dataset?
Sorry, perhaps I was not clear enough. No, you can't get OTU level assignments from a table summarised at phylum level.
What I was trying to say was, to do a phylogenetic tree you need to have the 16S rRNA gene sequences, and if you have them, you could process them to get an OTU table. Since you don't have 16S rRNA gene amplicon sequences for this data set, you can't do a phylogenetic tree neither obtain a table at OTU level.
There are several ways to obtain OTU tables, which rely on clustering the sequences at some identity threshold level. If you don't have the sequences, you can't cluster them, and, therefore you can't obtain a OTU abundance table.
António