function request: per sample filtering

[qiyunzhu]

Improvement Description
Hello qiime2 team, this is Qiyun Zhu from the Knight Lab. We (cc'ing @wasade and @tanaes) recently realized that filtering out low-abundance taxa on a per-sample base can effectively reduce table density, and improve the resolution of downstream beta diversity analysis, especially for those unweighted metrics. To my understanding, both QIIME1 and 2 do not have the function for per-sample filtering. Would it be a good idea to add this function?

Proposed Behavior
The implementation should be very simple. My solution (using the biom library) is:

th = 0.001  # threshold, which is a fraction

def filter_otus(data, id_, md):
    bound = data.sum() * th
    data[data < bound] = 0
    return data

table.transform(filter_otus, axis='sample')
table.remove_empty(axis='observation')

References
(Migrated from #396)

[mestaki]

Forum x reference