I want to construct a heat map at family level. I have 21 samples in total out of which nine belongs to subject 1 and twelve belongs to subject 2. I am able to generate a heat map but I want to filter out low frequency reads to obtain clear cluster which shows difference between two subjects. For example, Clostridiaceae are dominant in subject 1 samples having more than 1000 reads each but the same family is also present in subject 2 with less than 10 reads. Is it possible to filter the low abundant reads from subject 2 without affecting subject 1? For example, I need a heat map which gives a clear cluster to show difference between two subjects.
I would discourage that filtering. When you apply a feature-based filter, it needs to be applied uniformly to all samples, unless there’s a secondary motivating reason to apply it differently (i.e. I might filter a different set of features at different body sites or I might filter different features across sequencing runs if I thought there were different contaminants.) However, you should not filter your sequences differently here.
Have you considered using a Gneiss heatmap with co-occurance based clustering? If the difference is drive by samples, you should see it there.
I’d suggest looking at the qiime feature-table plugin for filtering options. Its got a far more comprehensive list of options can I can come up with off the top of my head.
