diff ramclustr_define_experiment_wrapper.py @ 0:42c2a25ff197 draft

planemo upload for repository https://github.com/RECETOX/galaxytools/tree/master/tools/ramclustr commit c321421a07bfdcc9ed423e9ed2ee794157984ba1
author recetox
date Wed, 29 Jun 2022 10:00:43 +0000
parents
children
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/ramclustr_define_experiment_wrapper.py	Wed Jun 29 10:00:43 2022 +0000
@@ -0,0 +1,90 @@
+import argparse
+import csv
+import sys
+
+
+ARGUMENTS = {
+    "Experiment": {"help": "Experiment name, no spaces.", "type": str},
+    "Species": {"help": "Genus species from which samples are derived.", "type": str},
+    "Sample": {"help": "Type of sample (e.g., serum, leaf).", "type": str},
+    "Contributer": {"help": "Your or your PI's name.", "type": str},
+    "platform": {"help": "Either GC-MS or LC-MS.", "type": str},
+    "chrominst": {"help": "Model of LC/GC instrument.", "type": str},
+    "msinst": {"help": "Model of MS instrument.", "type": str},
+    "column": {"help": "Column description.", "type": str},
+    "InletTemp": {"help": "Temperature of inlet.", "type": str},
+    "TransferTemp": {"help": "Temperature of GC to MS transfer line.", "type": str},
+    "mstype": {"help": "Type of mass spectrometer (one of QQQ, TOF, QTOF, Orbi, Q).", "type": str},
+    "msmode": {"help": "Positive or negative ion mode.", "type": str},
+    "ionization": {"help": "Ionization (EI, AP, or CI).", "type": str},
+    "msscanrange": {"help": "Scan range used for acquisition.", "type": str},
+    "scantime": {"help": "Time for each full scan spectrum (e.g. 0.2 seconds).", "type": float},
+    "deriv": {"help": "Derivitization (TMS, TBDMS, or None).", "type": str},
+    "MSlevs": {"help": "Number of levels of energy acquired - 1 typically.", "type": float},
+    "solvA": {"help": "Solvent A composition.", "type": str},
+    "solvB": {"help": "Solvent B composition.", "type": str},
+    "CE1": {"help": "Collision energy of acquisition of MS data.", "type": str},
+    "CE2": {"help": "Collision energy of acquisition for MSe/idMSMS data (when applicable).", "type": str},
+    "colgas": {"help": "Gas used for collisional dissociation.", "type": str},
+    "conevol": {"help": "Cone voltage used for acquisition.", "type": str}
+}
+
+
+def get_value(args, value, fill=True):
+    if fill:
+        return args[value]
+    return "fill"
+
+
+def write_gcms(csv_writer, args, fill=True):
+    csv_writer.writerow(["", "", ""])
+    csv_writer.writerow(["GC-MS", "", ""])
+
+    gcms = ["chrominst", "msinst", "column", "InletTemp", "TransferTemp", "mstype", "msmode", "ionization",
+            "msscanrange", "scantime", "deriv", "MSlevs"]
+    for arg in gcms:
+        csv_writer.writerow([arg, get_value(args, arg, fill=fill), ARGUMENTS[arg]["help"]])
+
+
+def write_lcms(csv_writer, args, fill=True):
+    csv_writer.writerow(["", "", ""])
+    csv_writer.writerow(["LC-MS", "", ""])
+
+    lcms = ["chrominst", "msinst", "column", "solvA", "solvB", "CE1", "CE2", "mstype", "msmode", "ionization", "colgas",
+            "msscanrange", "conevol", "MSlevs"]
+    for arg in lcms:
+        csv_writer.writerow([arg, get_value(args, arg, fill=fill), ARGUMENTS[arg]["help"]])
+
+
+def main(argv):
+    parser = argparse.ArgumentParser(description="RAMClustR define experiment.")
+
+    for arg in ARGUMENTS.keys():
+        parser.add_argument(f"--{arg}", type=ARGUMENTS[arg]["type"], help=ARGUMENTS[arg]["help"])
+
+    parser.add_argument("--output_file", type=str, help="Path to output csv file.")
+    args = vars(parser.parse_args())
+
+    with open(args["output_file"], "w") as csvfile:
+        csv_writer = csv.writer(csvfile, delimiter=',')
+
+        csv_writer.writerow(["parameter", "Value", "Description"])
+        csv_writer.writerow(["", "", ""])
+        csv_writer.writerow(["Experimental Design", "", ""])
+
+        general = ["Experiment", "Species", "Sample", "Contributer", "platform"]
+        for arg in general:
+            csv_writer.writerow([arg, args[arg], ARGUMENTS[arg]["help"]])
+
+        if args["platform"] == "GC-MS":
+            write_gcms(csv_writer, args)
+            write_lcms(csv_writer, args, fill=False)
+        elif args["platform"] == "LC-MS":
+            write_gcms(csv_writer, args, fill=False)
+            write_lcms(csv_writer, args)
+        else:
+            raise ValueError("Platform value incorrect, choose either GC-MS or LC-MS.")
+
+
+if __name__ == "__main__":
+    main(argv=sys.argv[1:])