Snakemake grouping output - snakemake

Question

I have a fairly complex requirement for the output of one of my rules.
I have samples in a CSV file (samples.csv) with nested information like so:
region,run_id,sample_id
A,1,150
A,3,111
A,3,145
A,3,153
B,4,105
B,4,107
I have them stored in my Snakefile as a pandas dataframe.
samples = pd.read_csv("samples.csv").set_index(
["region", "run_id", "sample_id"], drop=False
)
I have a rule demultiplex, which takes in a fastq file that is a combination of all the sample_id reads for a run_id in a region
rule demultiplex:
input:
"analysis/{region}/{run}/basecalled.fastq"
What this demultiplex rule does is take in this combined fastq file and returns a fastq file for each sample_id (barcode) within that run_id.
So where I am struggling here is how to get the output to map to the input.
What I effectively want in the running example is the demultiplex rule to execute once for every valid (region, run_id) pair, but produce multiple files. So the desired groupings would be something like this:
input:
"analysis/A/1/basecalled.fastq"
output:
["analysis/A/1/demultiplex/150.fastq"]
input:
"analysis/A/3/basecalled.fastq"
output:
["analysis/A/3/demultiplex/111.fastq",
"analysis/A/3/demultiplex/145.fastq",
"analysis/A/3/demultiplex/153.fastq"]
input:
"analysis/B/4/basecalled.fastq"
output:
["analysis/B/4/demultiplex/105.fastq",
"analysis/B/4/demultiplex/107.fastq"]
So the rule demultiplex in this example would run 3 times.

Answer 1

A solution could be to touch a file for each fastq file that needs to be demultiplexed. In the code below the assumption is that the path to the master fastq files can be reconstructed from the information in the sample sheet (which should be the case if I understand correctly):
import pandas as pd
samples = pd.read_csv("samples.csv").set_index(
["region", "run_id", "sample_id"], drop=False
)
fastq= samples[['region', 'run_id']].drop_duplicates()
rule all:
input:
expand('analysis/{region}/{run}/demultiplex/demux.done', zip, \
region= list(fastq['region']), run= list(fastq['run_id']))
rule demultiplex:
input:
"analysis/{region}/{run}/basecalled.fastq",
output:
touch('analysis/{region}/{run}/demultiplex/demux.done'),
shell:
r"""
demux {input}
"""