pyubero / microguilds Goto Github PK

If anyone wants to screen guilds in reduced contexts, this is, that not every single row the my master_tab.tsv will be classified in a context to visualize or to generate k-values, it would be useful to not getting this error in modify_mastertable.py:

ValueError: Length of values (2375) does not match length of index (91159)

I rewrite the modify_mastertable.py module to generate a master_tab with 4 context (instead of depth, latitude). I'm getting this error in guild_tensor_generate.py:

Found 4 contexts in gene subtable.
z1
z2
z3
z4

100%|██████████| 2064/2064 [00:01<00:00, 1746.92it/s]
Bivariate loglog regression results:
gamma = 0.6510109031387529
c = 0.2594626316691924
R2 = 0.8597118038702496

Gene: potF with R2=0.86
Data saved in kvalues_potF_Species_GTDB.tsv.
0%| | 0/2064 [00:00<?, ?it/s]
Traceback (most recent call last):

File "/Users/juanrivassantisteban/miniconda3/lib/python3.10/site-packages/spyder_kernels/py3compat.py", line 356, in compat_exec
exec(code, globals, locals)

File "/Users/juanrivassantisteban/Desktop/uGuilds-main/guild_tensors/guild_tensor_generate.py", line 129, in
gtutils.export_legacy(adu_table, _filepath, column="Diversity")

File "/Users/juanrivassantisteban/Desktop/uGuilds-main/guild_tensors/guild_tensor_utils.py", line 91, in export_legacy
assert sum(idx) == 1

AssertionError

It would be interesting to compare the sum of abundance values with the k-values. In order to do that in contexts with very different number of samples, we should add another column with Abudance / n samples.

Because sometimes different clusters represent different genes, so regressions need to be gene- and context-specific.

Sometimes, the number of represented colors does not match with the actual legend colors.

It would be nice that every implementation of several functions can be visualized at the same time. Things to consider:

• there's an obvious limitation of how many bars and colors fill in a radial plot...
• Are radial plots the best way to compare not a single guild pattern, but different guilds?

guild_tensor_generate.py

When trying to run a master_tab with only GTDB taxonomic assigments, guild_tensor_generate.py is displaying this error:

   File "/Users/juanrivassantisteban/Desktop/uGuilds-main/guild_tensors/guild_tensor_utils.py", line 132, in check_mastertable
    idc_null = np.argwhere(df['Species_SQM'].isnull().values)[:, 0]

  File "/Users/juanrivassantisteban/miniconda3/lib/python3.10/site-packages/pandas/core/frame.py", line 3807, in __getitem__
    indexer = self.columns.get_loc(key)

  File "/Users/juanrivassantisteban/miniconda3/lib/python3.10/site-packages/pandas/core/indexes/base.py", line 3804, in get_loc
    raise KeyError(key) from err

KeyError: 'Species_SQM''

And if you add an empty column called "Species_SQM", the error changes to this one:

  File "/Users/juanrivassantisteban/Desktop/uGuilds-main/guild_tensors/guild_tensor_utils.py", line 144, in check_mastertable
    df.at[jj, "Genus_SQM"] = "GTDB:"+df.loc[jj, "Genus_GTDB"]

TypeError: can only concatenate str (not "float") to str

It would be freaking good to compare distinct guild_tensors in the same plot.

Some concerns:

i) the k-values of different guilds can be in different magnitude orders. We should build a smart visualization!
ii) the resulting plot can represent an understanding-challenge. We should keep the things "easy".

It would be nice to add an argument for visualizing normalized abundances instead of k-values.

A major update can consist of a new module for computing, generating and displaying several suitable statistical tests for compositional analysis.

It would be nice to be able to order the appaerance of the implementations or clusters in the radial plots.

We need to compute diversity as the number of unique identifiers (rows) matching a value over 0 in the abundance column for any position in the k-tensor. Otherwise, diversity will always be the same (while abundance, of course, not) for every taxonomic level, regardless of the context.