Refactor symlink_files function and update

DesignIP class

seqnado/utils.py

@@ -448,9 +448,34 @@ class DesignIP(BaseModel):
     )
 
     @computed_field
+    @property
+    def sample_names_ip(self) -> List[str]:
+        sample_names = set()
+        for experiment in self.assays.values():
+            sample_names.add(experiment.ip_files.name)
+
+        return list(sample_names)
+
+    @property
+    def sample_names_control(self) -> List[str]:
+        sample_names = set()
+        for experiment in self.assays.values():
+            if experiment.control is not None:
+                sample_names.add(experiment.control_files.name)
+
+        return list(sample_names)
+
     @property
     def sample_names(self) -> List[str]:
-        return list(self.assays.keys())
+        return self.sample_names_ip + self.sample_names_control
+
+    @property
+    def ip_names(self) -> List[str]:
+        return list(set([experiment.ip for experiment in self.assays.values()]))
+
+    @property
+    def control_names(self) -> List[str]:
+        return list(set([experiment.control for experiment in self.assays.values()]))
 
     @computed_field
     @property
@@ -599,10 +624,16 @@ def symlink_files(
     r2_path_new = pathlib.Path(f"{output_dir}/{assay_name}_2.fastq.gz")
 
     if not r1_path_new.exists():
-        r1_path_new.symlink_to(assay.r1.path)
+        try:
+            r1_path_new.symlink_to(assay.r1.path.resolve())
+        except FileExistsError:
+            logger.warning(f"Symlink for {r1_path_new} already exists.")
 
     if assay.r2 and not r2_path_new.exists():
-        r2_path_new.symlink_to(assay.r2.path)
+        try:
+            r2_path_new.symlink_to(assay.r2.path.resolve())
+        except FileExistsError:
+            logger.warning(f"Symlink for {r2_path_new} already exists.")
 
 
 def symlink_fastq_files(
@@ -620,12 +651,14 @@ def symlink_fastq_files(
 
     elif isinstance(design, DesignIP):
         for experiment_name, experiment in design.assays.items():
-            for assay_name, assay in experiment.ip_files.items():
-                symlink_files(output_dir, assay, assay_name)
+            assay = experiment.ip_files
+            assay_name = assay.name
+            symlink_files(output_dir, assay, assay_name)
 
             if experiment.control_files:
-                for assay_name, assay in experiment.control_files.items():
-                    symlink_files(output_dir, assay, assay_name)
+                assay = experiment.control_files
+                assay_name = assay.name
+                symlink_files(output_dir, assay, assay_name)
 
 
 def define_output_files(
@@ -676,7 +709,7 @@ def define_output_files(
                 assay_output.extend(
                     expand(
                         "seqnado_output/peaks/{method}/{ip}.bed",
-                        ip=kwargs["sample_names_ip"],
+                        ip=kwargs["ip"],
                         method=peak_calling_method,
                     )
                 )

seqnado/workflow/rules/qc.smk

@@ -58,7 +58,7 @@ def get_fastqc_files(*args, **kwargs):
     fq_files = pathlib.Path("seqnado_output/fastqs").glob("*.fastq.gz")
     for fq_file in fq_files:
         fastqc_file = fastqc_dir / (fq_file.stem.replace(".fastq", "") + ".html")
-        fastqc_files.append(str(fq_file))
+        fastqc_files.append(str(fastqc_file))
 
     return fastqc_files
 

seqnado/workflow/snakefile_chip

@@ -4,41 +4,46 @@ import shutil
 from datetime import datetime
 import glob
 from snakemake.utils import min_version
-import seqnado.utils
+from seqnado.utils import DesignIP
+import seqnado.utils as utils
 import pandas as pd
 
+####################
+# Hardcoded config #
+####################
 ASSAY = "ChIP"
 configfile: "config_chip.yml"
 container: "library://asmith151/seqnado/seqnado_pipeline:latest"
 
-seqnado.utils.format_config_dict(config)
+####################
+# Experiment config #
+####################
 
-# Get experiment design
+# Load config
+utils.format_config_dict(config)
+
+# Generate design
 if os.path.exists(config["design"]):
-    # Expect columns - sample fq1 fq2 antibody control
-    FASTQ_SAMPLES = seqnado.utils.ChipseqFastqSamples(
-        pd.read_csv(config["design"], sep="[\s+,\t]", engine="python")
-    )
-    assert FASTQ_SAMPLES.design.shape[0] > 0, "No samples found in design file"
+    df = pd.read_csv(config["design"], sep="[\s+,\t]", engine="python")
+    DESIGN = DesignIP.from_dataframe(df)
+else:
+    DESIGN = DesignIP.from_directory(".")
 
-    for col in ["sample", "fq1", "fq2", "antibody", "control"]:
-        assert col in FASTQ_SAMPLES.design.columns, f"Design file must contain columns sample, fq1, fq2, antibody, control. Columns found: {FASTQ_SAMPLES.design.columns}"
+# Attempt to symlink the fastq files
+assert len(DESIGN.fastq_paths) > 0, "No fastq files found in the working directory or no design file provided."
+utils.symlink_fastq_files(DESIGN, output_dir="seqnado_output/fastqs")
 
-else:
-    # Use pattern matching to get samples
-    fq_files = list(seqnado.utils.get_fastq_files("."))
-    if fq_files:
-        FASTQ_SAMPLES = seqnado.utils.ChipseqFastqSamples.from_files(fq_files)
-    else:
-        raise ValueError("No FASTQ files found in the working directory")
-
-DESIGN = FASTQ_SAMPLES.design
-ANTIBODIES = FASTQ_SAMPLES.antibodies
-SAMPLE_NAMES = FASTQ_SAMPLES.sample_names_all
-SAMPLE_NAMES_IP = FASTQ_SAMPLES.sample_names_ip
-SAMPLE_NAMES_CONTROL = FASTQ_SAMPLES.sample_names_control
-SAMPLE_NAMES_PAIRED = FASTQ_SAMPLES.paired_ip_and_control
 
+
+
+# Define global variables
+SAMPLE_NAMES = DESIGN.sample_names
+SAMPLE_NAMES_IP = DESIGN.sample_names_ip
+SAMPLE_NAMES_CONTROL = DESIGN.sample_names_control
+IP = DESIGN.ip_names
+CONTROL = DESIGN.control_names
+
+# Load required rules
 include: "rules/qc.smk"
 include: "rules/fastq_trim.smk"
 include: "rules/align.smk"
@@ -49,13 +54,22 @@ include: "rules/heatmap.smk"
 include: "rules/hub.smk"
 
 
-ANALYSIS_OUTPUT = seqnado.utils.define_output_files(sample_names=SAMPLE_NAMES,
+# Define output files
+ANALYSIS_OUTPUT = seqnado.utils.define_output_files(
+                                            sample_names=SAMPLE_NAMES,
                                             assay=ASSAY,
-                                            sample_names_ip=SAMPLE_NAMES_IP,
+                                            ip=SAMPLE_NAMES_IP,
                                             **config
                                             )
 
 
+# Define wildcard constraints
+wildcard_constraints:
+    read = r"[12]",
+    sample = "|".join(SAMPLE_NAMES),
+    treatment= "|".join(SAMPLE_NAMES_IP),
+    control = "|".join(SAMPLE_NAMES_CONTROL),
+
 
 rule all:
     input: