Error occur during group_significance.py using rarefied BIOM data

[Yoko Nagai]

Hello,

I got an error using QIIME 1.9.1 group_significance.py for rarefied BIOM data, whether original BIOM data went without any problem or error.

I used unziped BIOM data that is created by core_diversity_analsyses.py

Command:

group_significance.py -i unzip_table_even5000_from_core_diversity_analysis.biom -m map.txt -c Group -s mann_whitney_u -o out.txt

Error:

Traceback (most recent call last):

  File "/usr/local/bin/group_significance.py", line 344, in <module>

    main()

  File "/usr/local/bin/group_significance.py", line 326, in main

    GROUP_TEST_CHOICES, int(opts.permutations))

  File "/usr/local/lib/python2.7/dist-packages/qiime/otu_significance.py", line 147, in run_group_significance_test

    test_stat, pval = test_choices[test](row[0], row[1])

  File "/usr/local/lib/python2.7/dist-packages/qiime/stats.py", line 1882, in mw_t

    u, pval = mannwhitneyu(x, y, continuity)

  File "/usr/local/lib/python2.7/dist-packages/scipy/stats/stats.py", line 4092, in mannwhitneyu

    raise ValueError('All numbers are identical in amannwhitneyu')

ValueError: All numbers are identical in amannwhitneyu

Do I need to re-create rarefied dataset?

[Colin Brislawn]

Hello Yoko,

Thank you for posting the full command and error.

Based on this error "All numbers are identical in amannwhitneyu" it looks like the stat test cannot be performed because all the numbers are identical!

The group_significance.py script should work just fine on normalized data, but the results can be strange and make errors if the rarefaction depth is too low or if you have too few samples. Also, rarefaction removes samples under your rarefaction level (so samples under 5000 reads would be dropped), and this is could also make you have two few samples.

How many samples are in that even5000 table? Does this script work if you rarify to 1000 reads per sample (instead of 5000)?

Thanks,

Colin