added classifier.check_csv

Author: HannesK

classifier.py

@@ -49,14 +49,24 @@ def csv2rows(FnameCSV):
 
     with open(FnameCSV, 'rb') as f:
         reader = csv.reader(f, delimiter=",")
-        header = reader.next()
-        header = [x.strip() for x in header]
+        for x in reader:
+            if not x:continue
+            header = x
+            header = [x.strip() for x in header]
+            break
+        
         miRNAs = [x for x in header if not x in ["ID", "Annots"]]
 
+        IDs = set([])
         rows = []
         for x in reader:
             if not x: continue
+            if not x[0].strip(): continue
             newrow = dict(zip(header,[y.strip() for y in x]))
+            if newrow["ID"] in IDs:
+                print "\n***ERROR: row IDs must be unique, found duplicate (%s)."%newrow["ID"]
+                raise Exception
+            IDs.add(newrow["ID"])
             rows.append(newrow)
 
     return miRNAs, rows
@@ -368,9 +378,6 @@ def function(SampleDict):
 
     return function
 
-        
-        
-
 
 
 def check_classifier(FnameCSV, GateInputs):
@@ -407,8 +414,37 @@ def check_classifier(FnameCSV, GateInputs):
         print " result = classifier and data are consistent"
 
 
-       
-       
+
+def check_csv(FnameCSV):
+    """
+    counts how many miRNAs are constant across all samples, and
+    checks if there are inconsistencies in the data (identical miRNA profile but different annotation)    
+    """
+
+    print "\n--- check_csv"
+
+    miRNAs, rows = csv2rows(FnameCSV)
+    print " miRNAs: ", len(miRNAs)
+    print " samples:", len(rows)
+    
+    inconsistencies = []
+    seen = []
+    for x in rows:
+        for y in seen:
+            if all(x[rna]==y[rna] for rna in miRNAs):
+                if x["Annots"]!=y["Annots"]:
+                    inconsistencies.append(x["ID"])
+        seen.append(x)
+
+    constants = []
+    for rna in miRNAs:
+        value = rows[0][rna]
+        if all(x[rna]==value for x in rows):
+            constants.append(rna)
+    
+    print " inconsistencies (%i): %s"%(len(inconsistencies),",".join(inconsistencies) or "-")
+    print " constants (%i): %s"%(len(constants),",".join(constants) or "-")
+
 
 def mat2csv(FnameMAT, Threshold):
 	"""

readme.md

@@ -207,7 +207,11 @@ The output informs you of all encountered inconsistencies (malfunctions):
  result = 1 inconsistencies set(['2'])
 ```
 
-
+#### check csv data
+To check whether there are _inconsistencies_ or _constants_ in the data call the function `classifier.check_csv`.
+It lists all miRNAs that are constant across all samples, i.e., always _high_ or always _low_,
+and checks if there are profiles that are identical in terms of the miRNA profile but differ w.r.t. to the annotation.
+If there are inconsistencies then there is no classifier.
 
 
 ## Files

toy.csv

@@ -1,7 +1,8 @@
-ID, Annots, g1, g2, g3
-1,  0,      1,  1,  0
-2,  0,      0,  0,  1
 
+ID, Annots, g1, g2, g3 
+      
+1,  0,      1,  1,  0   
+2,  0,      0,  0,  1
 3,  1,      0,  1,  0
 
 

toy.py

@@ -43,7 +43,7 @@
 import classifier
 
 if __name__=="__main__":
-    if 1 :
+    if 1:
         classifier.csv2asp(
             FnameCSV,
             FnameASP,
@@ -53,18 +53,22 @@
             EfficiencyConstraint,
             OptimizationStrategy)
 
-    if 1 :
+    if 1:
         GateInputs = "gate_input(1,positive,g2) gate_input(2,positive,g3) gate_input(2,negative,g1)"
         classifier.check_classifier(FnameCSV, GateInputs)
 
-    if 1 :
+    if 1:
         GateInputs = "gate_input(1,positive,g2) gate_input(2,positive,g3) gate_input(2,negative,g1)"
         FnamePDF = "toy.pdf"
         classifier.gateinputs2pdf(FnamePDF, GateInputs)
 
-    if 0 : 
-    	  FnameMAT = "toy.mat"
-    	  Threshold = 250
-    	  classifier.mat2csv(FnameMAT, Threshold)
+    if 0:
+        FnameMAT = "toy.mat"
+        Threshold = 250
+        classifier.mat2csv(FnameMAT, Threshold)
+
+    if 1:
+        classifier.check_csv(FnameCSV)
+