Suppose I have the following input FASTQ files:

 $ ls  -1 inputs/ 
CM0009619-ABB_L01_Read1_Sample_Library_XY102.fastq.gz
CM0009619-ABB_L01_Read1_Sample_Library_XY110.fastq.gz
CM0009619-ABB_L01_Read1_Sample_Library_XY84.fastq.gz
CM0009619-ABB_L01_Read2_Sample_Library_XY102.fastq.gz
CM0009619-ABB_L01_Read2_Sample_Library_XY110.fastq.gz
CM0009619-ABB_L01_Read2_Sample_Library_XY84.fastq.gz
CM0009619-ABB_L02_Read1_Sample_Library_XY84.fastq.gz
CM0009619-ABB_L02_Read1_Sample_Library_XY88.fastq.gz
CM0009619-ABB_L02_Read2_Sample_Library_XY84.fastq.gz
CM0009619-ABB_L02_Read2_Sample_Library_XY88.fastq.gz

There are three variables: the lane identifier (L0*), read identifier (Read*) and the sample identifier (XY*).

All samples have both Read1 and Read2. Most samples belong to just 1 lane, but they can belong to both. Therefore, the combination of {sample} and {lane} must be conditional.

I’m trying to figure this out using a simple test workflow but the solution has eluded me for many hours

SAMPLES={"XY84": (1, 2),
         "XY88": 2,
         "XY102": 1,
         "XY110": 1
}

rule all:
  input:
    expand("inputs/CM0009619-ABB_L0{lane}_Read{read}_Sample_Library_{sample}.fastq.gz", sample=SAMPLES.keys(), read=[1,2], lane=SAMPLES[sample] )

rule copy_files:
  input:
    expand("inputs/CM0009619-ABB_L0{lane}_Read{read}_Sample_Library_{sample}.fastq.gz", sample=SAMPLES.keys(), read=[1,2], lane=SAMPLES[sample] )   
  output:
    expand("outputs/CM0009619-ABB_L0{lane}_Read{read}_Sample_Library_{sample}.fastq.gz", sample=SAMPLES.keys(), read=[1,2], lane=SAMPLES[sample] )
  shell:
    "cp {input} {output}"   

However I get the error:

$ snakemake --snakefile Snakefile --cores 1 
NameError in line 11 of /path/to/smktests/Snakefile:
name 'sample' is not defined
  File "/path/to/smktests/Snakefile", line 11, in <module>

Clearly I cannot use the sample value in the expansion function as the key to the SAMPLES dictionary when trying to get the appropriate lane value. However, simply using lane=[1,2] also does not work because this generates some invalid combinations.

Is there a simple way to have the lane value depend on the sample value? Is Snakemake inappropriate for this scenario?

I’m using v5.26.1 of Snakemake but could consider a more recent version if it gets me to a solution.