Error while reconstructing a table via sidle plugin

Hi everybody,

I´m analyzing my first microbial dataset and ran into an issue trying to reconstruct a table using the sidle plugin command "qiime sidle reconstruct-counts".

I followed the very clear and useful sidle tutorial! :slight_smile:
I have demultiplexed my dataset containing reads from the bacterial V45 and V6 region succesfully and everything worked well until I encountered an error after using this code:

qiime sidle reconstruct-counts \
--p-region V45 \
--i-regional-alignment ${alignment_V45} \
--i-regional-table ${table_V45} \
--p-region V6 \
--i-regional-alignment ${alignment_V6} \
--i-regional-table ${table_V6} \
--i-database-map ${bacterial_map} \
--i-database-summary ${bacterial_summary} \
--p-n-workers ${cores} \
--o-reconstructed-table ${table_bacterial_reconst} \
--verbose

The error was:

There are 32 samples with fewer than 1000 total sequences. Please check your minimum counts and make sure your representative sequences are aligned with the database.
  1. I assume filtering out the 32 samples with fewer than 1000 total sequences would solve the error partly but I don´t know which file I would have to filter and could not find a method to filter by sequence counts.
  2. Could you tell me how or in which file I have to check the minimum counts since the visualized "table_V45" and "table_V6" only show frequencies?
  3. How do I check if my representative sequences are aligned with the database? Is there a way to visualize the region specific alignments?

I am kind of lost on what file or parameter used in previous code, which was all according to the tutorial, could cause the problem or help solve the error.

Thanks in advance.
Kind regards,
Lea

1 Like

Hi @lea.k,

I'm glad to see people using Sidle!

This is an issue where we've gone back and forth on whether to error or let things pass. There were a lot of people frustrated they weren't getting an error and just losing a bunch of samples (which was the original implementation).

I have good news and bad news for solving this. The good news is that there is code. The bad news is that while it's been tested, it hasn't been reviewed and there's no documentation. (The documentation is half complete, but unfortunately, Sidle is currently lower priority and I have less free time than I did a few years ago.) The branch is called "accounting", and I think you can install the code from that branch with

pip install git+https://github.com/jwdebelius/[email protected]

(Although you may need to do a little bit more reserach into how to install a branch, but I think this should work. :crossed_fingers: )

The function that will let you count yoru features is called track-aligned-counts and it takes your regional alignment(s), feature table(s) and produces an artifact that you can then summarize wtih qiime metadata tabulate. (Hopefully the help documentation explains it well). If it works as intended, it will give you a list of the aligned sequencing depths, and you could use that to filter your tables.

I typically summarize the regional alignment artifact (again qiime metadata tabulate and look at how many reads get aligned per region. If I dont see enough, I tend to relax the parameters slightly. You're working in a region that's got fairly good database coverage, so you may want to consider allowing ASVs that don't match perfectly to align.

Sorry I dont have a perfect, out of the box solution.

Best,
Justine

1 Like