Suggestions

[corneliusroemer]

I tried to run the benchmarks but struggled. I suggest to make it more reproducible to:

• provide a conda environment.yml to install the required software, e.g. I couldn't find bambu, nanocount and bustools in bioconda nor conda-forge

Here's what I got as a start:

name: lr_quant_benchmark
channels:
- conda-forge
- bioconda
- nodefaults
dependencies:
- kallisto
- oarfish
- python
- snakemake

• Have one main Snakefile at top level of directory that's called Snakefile
• Download data as part of the workflow, not via separate instructions in the README
• provide input/output dirs that work straight after cloning and not use things like /mnt, so that one can run the analysis straightaway without having to change things

[rob-p]

Hi @corneliusroemer,

Thanks for the pointers. These are great suggestions. I just spoke with @NPSDC and @zzare-umd yesterday about having a top-level Snakemake file. I think that's the way to go.

I'm not certain that it's ideal to assume that all tools are present directly in the path, as users may want to evaluate different versions of tools and provide custom paths to them (the replicable results are expected, of course, only for the versions we list here).

As I can see that you're already addressing upstream, the newest version of NanoCount can currently only be obtained from their GitHub repository and installed from source (and not via bioconda).

We will continue to try to make the pipeline simpler to run, and we'd be happy to welcome discussion on any PRs you might make, and include them if they make sense!

Best,
Rob

[corneliusroemer]

I'm not certain that it's ideal to assume that all tools are present directly in the path, as users may want to evaluate different versions of tools and provide custom paths to them (the replicable results are expected, of course, only for the versions we list here).

People can customize their environment if they want to - but offering a default is good practice to simplify testing.

As I can see that you're already addressing upstream, the newest version of NanoCount can currently only be obtained from their GitHub repository and installed from source (and not via bioconda).

One can pip+git install, see PR #2 ;)