📝 Modify documentation for benchmark.rst

[Sidduppal]

📝 Add docs for classification and clustering-classification 📝 Add docs to run multiple results of the same community at once 🔥 Remove code to aggregate results

[chasemc]

I think it may be helpful to throw an example into Jupyter notebook and link to it (or https://docs.readthedocs.io/en/stable/guides/jupyter.html) Especially since the tutorial jumps around between running things on command line and python

[chasemc]

I think the document would benefit from a bit of rearrangement Including: Autometa Test Datasets should be combined with Downloading Test Datasets and should be in the same location as Downloading Test Datasets. Titles:

• Benchmarking Datasets
  • Download Datasets
  • Explanation of Datasets
    • Synthetic Communities
    • Simulated Communities
    • Generating New Simulated Communities

Move the heading Example benchmarking with simulated communities to directly before the heading Benchmark clustering

---

Make the examples for each type of clustering clearer by presenting them in the same context :

• Benchmark clustering is presented in the context of for-looping over community sizes
• Technically Benchmark classification is presented in the context of for-looping over community sizes but the for-loop isn't shown
• Benchmark clustering-classification is presented in the context of running only on a single community size

The example should be simple, so I would suggest presenting only with the context of running on a single sample/community size.

An "Advanced" section could discuss how to handle running on multiple samples.

---

Generally, the commands/code need more description. As it stands now it's still unclear what commands to run if I want to benchmark (Is following the section Benchmark clustering-classification the same as running Benchmark clustering and Benchmark classification separately?) Possibly could be done with more descriptive headings (e.g. Benchmark clustering becomes Benchmark Binning Results; Benchmark classification becomes Benchmark taxonomic assignments), etc

Aggregate results across simulated communities seems to be just data handling? If just data-handling, it makes the documentation less clear; remove or place in an "Advanced" section or similar at end.

Flags such as --output-long and --output-classification-reports aren't described/defined Inputs aren't described, e.g. how is the input file for --predictions, --reference, etc supposed to be structured?

The documentation is missing discussion of what autometa-benchmark actually does, and the results it produces.

[evanroyrees]

Closing this for now. Please submit a new PR from a new KwanLab/Autometa branch with any changes still necessary.

Addressed a few comments in https://github.com/KwanLab/Autometa/pull/215