3

I'm using the python API for the biom-format package. I read in two files in biom format and merged them, like so:

In [1]: import biom

In [2]: a = biom.load_table('./table-a.biom')

In [3]: b = biom.load_table('./table-b.biom')

In [4]: a
Out[4]: 51116 x 46 <class 'biom.table.Table'> with 185077 nonzero entries (7% dense)

In [5]: b
Out[5]: 37479 x 26 <class 'biom.table.Table'> with 100124 nonzero entries (10% dense)

In [6]: c = a.merge(b)

In [7]: c
Out[7]: 67022 x 47 <class 'biom.table.Table'> with 254555 nonzero entries (8% dense)

How do I now export this biom object (c) as a biom file? There are a few methods available, eg, .to_tsv(), but I don't see a way to save this object in biom format.

Devon Ryan
  • 19,602
  • 2
  • 29
  • 60
John
  • 147
  • 4
  • Are you using the biomformat package of bioconductor through Python? Or the biom-format package? Anyway in R the relevant function is write_biom so it might be the same in the python package – llrs Oct 25 '17 at 12:42

2 Answers2

4

You can use the functions to_hdf5 and to_json from the biom-format package:

to_hdf5

with biom.util.biom_open('output.biom', 'w') as f:
    merged_table.to_hdf5(f, "example")

to_json

with open('output.txt', 'w') as f:
    f.write(merged_table.to_json("example"))
user3648523
  • 41
  • 3
1

There is a function in PiCRUST you can use

from picrust.util import write_biom_table

write_biom_table(c,'table-c.biom')

Note, however, that depending if you want json or hdf5 format and if your metadata are complex, you may need to use additional arguments. These are the arguments of the function:

def write_biom_table(biom_table, biom_table_fp, compress=True,
                 write_hdf5=HAVE_H5PY, format_fs=None):
"""Writes a BIOM table to the specified filepath

Parameters
----------
biom_table : biom.Table
    The table object to write out
biom_table_fp : str
    The path to the output file
compress : bool, optional
    Defaults to ``True``. If True, built-in compression on the output HDF5
    file will be enabled. This option is only relevant if ``write_hdf5`` is
    ``True``.
write_hdf5 : bool, optional
    Defaults to ``True`` if H5PY is installed and to ``False`` if H5PY is
    not installed. If ``True`` the output biom table will be written as an
    HDF5 binary file, otherwise it will be a JSON string.
format_fs : dict, optional
    Formatting functions to be passed to `Table.to_hdf5`

Notes
-----
This code was adapted from QIIME 1.9
"""

I have some problems to perform a simple operation and write the output of biom data generated with picrust, it might be helpful as an example of how to use the arguments.

Alf Pascu
  • 111
  • 1