I have a feature table I recived from collaborator, using Qiime1 to process the sequences.
It looks like feature tables usually look, with first coulmn containing feature information and each coulmn is a sample with counts:
One of the delightfully weird tricks about the python API is that the pandas representation of a feature table uses samples as rows and features as columns, but the biom Table representation uses features as row and samples as columns.
QIIME 1 was based on a biom format exclusively with a biom export. My suspecion is the at the table was exported using the biom convert functionality which didn't do transposition.
So, yeah, transpose your table or convert the text (pandas) table to a biom.Table object before importing.