Working with sourcetracking without Source

[shashankx]

Hi,
I have human skin microbiome data (ASV table). Can we use the Sourcetracking tool to see how much percentage of taxa coming from let's say soil?
As I can see from the tutorial, in an abundance table, it should contain both Sink and Source. Here in my case, I only have Sink (input is skin samples), however, I don't have a Source.

Is there a reference database by which I can use for my Skin microbiome data and look at the possible source? If not what is an alternative for that?
Thank you

[AdeBC]

Hi, I think the reference database you've mentioned is more likely to be contained in a supervised-learning method, cause supervised-learning methods often need to be trained on pre-labeled dataset (just like reference database). But for unsupervised-learning methods (like SourceTracker, FEAST), I don't think that's necessary.
My advice is that retrieving source data from MGnify, or you can try our new supervised-methods (see my repo "ONN4MDM"), but unfortunately it haven't been finished right now. So maybe you need that later, good luck!

[NeginValizadegan]

Hi,
I have human skin microbiome data (ASV table). Can we use the Sourcetracking tool to see how much percentage of taxa coming from let's say soil?
As I can see from the tutorial, in an abundance table, it should contain both Sink and Source. Here in my case, I only have Sink (input is skin samples), however, I don't have a Source.

Is there a reference database by which I can use for my Skin microbiome data and look at the possible source? If not what is an alternative for that?
Thank you

Hi @shashankx

Yes, as @AdeBC mentioned, you would have to download and analyze data from online sources such as qiita or MGnify and then use as the source. I think though, it might be best if you use data that are as similar as possible. For example, if you use amplicon data, you might want to select same amplicon regions, same platform if possible, and same analysis. Usually, close-reference OTU picking helps with some discrepancy issues, but you might still not see some of the similarities due to these methodological differences in your source and sink.