Is necessary an abundance table normalization/standarization when using K-means?

Good afternoon everyone,

Even if I am quite new on it, I'm conducting a data analysis project and one of the data sources is a microbiome abundance table.

One of the steps we would like to apply to data is a clustering analysis. Having already the relative abundance table, normalization of data would not be necessary, right? Should I apply any normalization method before?

Thank you so much in advanced :slight_smile:

Hi @MiriamGorostidi, I don't have experience k-means on microbiome data, so I'm not certain. We also don't have direct support for it in QIIME 2.

We do have support for UPGMA and neighbor-joining clustering of samples (see beta-rarefaction), which allow you to use microbiome-relevant distance metrics (e.g., unifrac or bray-curtis). I would recommend starting there if using k-means itself isn't a requirement.

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.