Advice on feature table filtering

General Discussion
,
1

Hello QIIME 2 community,

I'm working on analyzing microbial community data and need some advice on two aspects:

  1. Feature Table Filtering - Minimum Frequency Threshold
    I want to retain the most relevant features while filtering out low-abundance features, but I’m unsure which approach would be ideal. I've seen different suggestions, like setting the minimum frequency threshold to 0.01% of the maximum feature frequency or using the mean feature frequency divided by 0.001. In your experience, which approach would retain the best-quality features without introducing too much noise? Are there other strategies I should consider?
  2. Denoising Parameters - Balancing Feature Count and Abundance
    I've noticed that tweaking denoising parameters impacts my feature abundance and feature count quite significantly. With one set of parameters, I get high feature counts but lower abundance, while a different set leads to higher abundance but lower feature counts. Could anyone recommend an approach or specific parameters to help me strike a balance between the two?
1 Like
2

Hello Vanshikia,

Welcome to the forums! :qiime2:

These are great questions. I'm glad you are thinking carefully about parameters and how to produce the highest-quality data.

While there are papers on these topics, I think it's important to do your own benchmark so you can see for yourself. This way, you don't have to decide which advice is best; you can see what works best for your samples!

To do this, you need positive controls. Did you include axenic samples or samples with a known composition on your sequencing run?