list of taxa or features ??

[balathl]

Hi,

Is it possible to obtain the list of features explaining the prediction of source to sink in actual samples?

In other words, can we get the OTUIDs that were transfered from source to target or sink samples shown in the pie charts in Sourcetracker outputs? Also, can i find somewhere in the output on the % of source explained in pie charts or bar charts in txt or excel format??

Bala

[CamEJ]

Was looking for the same thing when using the R version of sourcetracker-1.0.1 but was never able to get the 'full results' folder containing list of features that was meant to come with the analysis.
After searching, I found that Sourcetracker2 has this feature, and it's called 'per_sink_feature_assignments' as found here :
https://github.com/biota/sourcetracker2
It isn't available in R it seems, so just used command line.

I followed the conda installation instructions (on a linux, as wouldn't work on my mac) and then used the command line instructions in the last of the Usage Examples:
_

"Calculate the proportion of each source in each sink, with 5 jobs. Write the per sink feature tables (what SourceTracker 1 called 'full output'). These are feature by sample table indicating the origin source of each sequence (each count of a feature)."
_
where the command I used was:
sourcetracker2 -i otu_table.biom.txt -m map.txt -o OutputFolder/ --jobs 15 --per_sink_feature_assignments

[theo-allnutt-bioinformatics]

Thanks CamEJ.. I needed to do this too. Although this diagram is shown in the Nature Methods paper, there is no example of how to get these results in example.r (I don't know how these things get past reviewers!). I will try sourcetracker2.

[casett]

Other than using sourcetracker2, was anyone able to get this to work in R? I was able to run with full.results = TRUE in R which says it "return full draws from gibbs sampling (all assignments of all taxa)" - but don't know how to parse it?

[FFoQS90]

I was having the same problem and I think I found a solution. I am still pretty new to R so there might be a better way to code this but I hope this helps:

#Contribution of individual taxa
#The "results" object has a four-dimensional "full.results" array that is Num trials x Num envs x Num OTUs x Num samples. So to see the number of observations of OTU 99 assigned to all sources in sample 3 for all trials/restarts, you would enter:

results$full.results[,,99,3]

#To get a matrix of OTU x Probability(OTU came from source) across all samples
asv.env.probs <- t(apply(results$full.results,c(2,3),mean))

asv.env.probs <- asv.env.probs[rowSums(asv.env.probs) > 0,]

asv.env.probs <- sweep(asv.env.probs,1,rowSums(asv.env.probs),'/')

#Add correct rownames and columnnames and add taxaID
taxatable <- as(tax_table(physeq_no_cont),"matrix")
taxatable <- as.data.frame(taxatable)
row.names(asv.env.probs) <- row.names(taxatable)
colnames(asv.env.probs) <- c("Anal Swab","Classification (Gloves)","Classification (Railing)","Cooling chamber (Wall)","Evisceration (Aprons)","Evisceration (Gloves)","Evisceration (Knifes)","Lock (Hands)","Lock (Shoes)","Lock brush (Clean area)","Lock liquid (Clean area)","Polishing tunnel (Nozzles)","Polishing tunnel (Water)","Polishing tunnel (Whips)","Splitting saw","Sticking","Truck (Gloves)","Truck (Wall)","Unknown")
asv.env.probs <- as.data.frame(asv.env.probs)

taxa_sourcetracker <- cbind(asv.env.probs,taxatable)
taxa_sourcetracker <- melt(taxa_sourcetracker)
taxa_sourcetracker <- taxa_sourcetracker[!(is.na(taxa_sourcetracker$value) | taxa_sourcetracker$value==""), ]