decontam applicable for curating data from TSC

[den-dal]

Hi!

I'm currently using the prevalence method with decontam to remove contaminants from my negative controls, and it's been very helpful. However, I’ve been handling tag-switching controls (TSC) manually, and I was wondering if there’s a way to implement a more automated approach, similar to what decontam offers, for filtering OTUs that appear in my TSC samples.

For context, I labeled my samples during library preparation using Nextera indexes, and I deliberately left some index combinations unused to detect tag-switching events. Would you happen to know of any methods or tools that could help automate the filtering of TSC-related artifacts? I'd greatly appreciate any advice or suggestions you might have.

Thanks in advance for your time and help!

[benjjneb]

I'm not completely up to date with the tooling on cross-contamination (which you are correct, decontam was not intended to remove). Note that this is an issue for sequencing broadly, not just amplicon sequencing or metagenomics. A couple references that might help are:

https://www.nature.com/articles/nmeth.4666
https://www.biorxiv.org/content/10.1101/2024.10.21.619353v2.abstract
https://doi.org/10.1111/1755-0998.13009

For myself, I am wary of "read subtraction" approaches generally -- they can do more harm than good. In the experiment we performed where we really cared about accurate ID of low abundance variants we used our "TSC"s to set a threshold over which we believed the taxa really was in the sample, and treated with more skepticism taxa below that level (which was fairly low in our experiment, ~0.01% if I remember correctly).

[den-dal]

Ok! Thank you very much for the references and your feedback!

[hjarnek]

@den-dal,

You could try UNCROSS2 from USEARCH v11 (otutab_xtalk command). Word of caution from the developer though, that de-novo estimation of cross-contamination rates is quite rough, and the program generally removes too much rather than too little, but if you have mock control samples, you can calibrate against those for better accuracy.