Corrected trap dataset handling

workflow/config/main.yaml

@@ -16,6 +16,8 @@ fast_min_q: 8 # empty string or 0 to disable
 pod5_dataset: "nomiss_96BC_P2I_SUP_2026"
 pod5_dataset_size : 293
 pod5_stride : 12 # Use 1 in x pod5 files from the ONT NO-MISS dataset to reduce dataset size / compute requirements
+trap_dataset_size: 1717
+trap_pod5_stride: 72
 trap_read_limit : '-n 500000'
 
 # custom ML

workflow/rules/baseline_pipeline.smk

@@ -32,33 +32,50 @@ def get_batch_range(wildcards):
     if wildcards.dataset == 'nomiss':
         return range(1,config['pod5_dataset_size']+1,config['pod5_stride'])
     if wildcards.dataset == 'trap':
-        return [0]
+        return [f'batch{x}' for x in range(1,config['trap_dataset_size']+1,config['trap_pod5_stride'])]
 
 def get_read_limit(wildcards):
     if wildcards.dataset == 'trap':
         return config["trap_read_limit"]
     return ''
 
+wildcard_constraints:
+    batch=r"(batch)?\d+",
+    model=r'hac|fast'
+
 rule download_nomiss_pod5:
     output:
         '../data/raw_pod5/nomiss/PBK98658_853a956f_57f83f46_{batch}.pod5'
     threads: 4
-    wildcard_constraints:
+    
-        batch=r"\d+"
     shell:
         """
         aws s3 cp --no-sign-request s3://ont-open-data/nomiss_96BC_P2I_SUP_2026/raw/pod5/PBK98658_853a956f_57f83f46_{wildcards.batch}.pod5 {output}
         """
 
-rule download_trap_pod5:
+rule download_trap_tar:
     output:
-        '../data/raw_pod5/trap/ATCC_25922_202309_0.pod5'
+        temp(directory('../data/raw_pod5/trap/all'))
     threads: 32
-    wildcard_constraints:
+    params:
-        batch=r'\d+'
+        tmp='/tmp/trap_all.tar'
     shell:
         """
-        curl -L --http1.1 -A "Mozilla/5.0" "https://api.figshare.com/v2/file/download/45408628" -o {output}
+        curl -L --http1.1 -A "Mozilla/5.0" "https://api.figshare.com/v2/file/download/45408628" -o {params.tmp}
+        tar xf --wildcards -C {output} {params.tmp} 
+        rm {params.tmp}
+        """
+rule extract_trap_pod5:
+    input:
+        '../data/raw_pod5/trap/all'
+    output:
+        '../data/raw_pod5/trap/ATCC_25922_202309_{batch}.pod5'
+    threads: 1
+    params:
+        input_pod5='{input[0]}/ATCC_25922__202309/ATCC_25922__202309_{batch}.pod5'
+    shell:
+        """
+        mv {params.input_pod5} {output}
         """
 rule basecall_pod5:
     input:
@@ -75,9 +92,6 @@ rule basecall_pod5:
         min_qscore=get_qscore,
         dorado_model=get_model_name,
         n_reads=get_read_limit
-    wildcard_constraints:
-        batch=r"\d+",
-        model="hac|fast"
     shell:
         """
         dorado basecaller --models-directory {config[dorado_model_dir]} {params.n_reads} --emit-fastq {params.benchmarking} {params.min_qscore} {params.dorado_model} {input.pod5} > {output}