Source Code for Module ML.ModelPackage.Packager

  1  # 
  2  # Copyright (C) 2002 Greg Landrum and Rational Discovery LLC 
  3  # All rights are reserved. 
  4  # 
  5   
  6   
  7  import exceptions 
  8 -class DescriptorCalculationError(exceptions.Exception): 
  9    """ used to signal problems generating descriptor values """ 
 10    pass 
 11 -class ClassificationError(exceptions.Exception): 
 12    """ used to signal problems generating predictions """ 
 13    pass 
 14   
 15 -class ModelPackage(object): 
 16    """ a container class to package a composite model with a descriptor 
 17    calculator so that objects needing predictions (compounds, molecules, etc.) 
 18    can be passed directly in without worrying about generating descriptors 
 19   
 20    """ 
 21 -  def __init__(self,descCalc=None,model=None,dataSet=None,notes=''): 
 22      self._descCalc = descCalc 
 23      self._model = model 
 24      self._notes = notes 
 25      self._dataSet = dataSet 
 26      self._initialized = 0 
 27      self._supplementalData = [] 
 28       
 29 -  def SetCalculator(self,calc): 
 30      self._descCalc = calc 
 31 -  def GetCalculator(self): 
 32      return self._descCalc 
 33   
 34 -  def SetModel(self,model): 
 35      self._model = model 
 36 -  def GetModel(self): 
 37      return self._model 
 38   
 39 -  def SetDataset(self,data): 
 40      self._dataSet = data 
 41 -  def GetDataset(self): 
 42      return self._dataSet 
 43   
 44 -  def SetNotes(self,notes): 
 45      self._notes = notes 
 46 -  def GetNotes(self): 
 47      return self._notes 
 48   
 49 -  def SetSupplementalData(self,suppD): 
 50      self._supplementalData = suppD 
 51 -  def GetSupplementalData(self): 
 52      if not hasattr(self,'_supplementalData'): 
 53        self._supplementalData = [] 
 54      return self._supplementalData 
 55 -  def AddSupplementalData(self,data): 
 56      if not hasattr(self,'_supplementalData'): 
 57        self._supplementalData = [] 
 58      self._supplementalData.append(data) 
 59     
 60 -  def Classify(self,obj,label='',threshold=0): 
 61      if not self._initialized: 
 62        self.Init() 
 63      try: 
 64        descs = self._descCalc.CalcDescriptors(obj) 
 65      except: 
 66        raise DescriptorCalculationError,'problems encountered generating descriptors' 
 67   
 68      argVect = [label]+list(descs)+[0] 
 69      try: 
 70        res = self._model.ClassifyExample(argVect,threshold=threshold,appendExample=0) 
 71      except: 
 72        raise ClassificationError,'problems encountered generating prediction' 
 73   
 74      return res 
 75   
 76 -  def Init(self): 
 77      if self._model is None or self._descCalc is None: 
 78        return 
 79       
 80      nms = self._model.GetDescriptorNames() 
 81      lbl = nms[0] 
 82      act = nms[-1] 
 83      descs = self._descCalc.GetDescriptorNames() 
 84      order = [lbl] + list(descs) + [act] 
 85      self._model.SetInputOrder(order) 
 86       
 87      self._initialized = 1 
 88   
 89  if __name__=='__main__': 
 90    from Chem import * 
 91    import cPickle 
 92    from ML.ModelPackage import Packager 
 93     
 94    calc = cPickle.load(open('test_data/Jan9_build3_calc.dsc','rb')) 
 95    model = cPickle.load(open('test_data/Jan9_build3_model.pkl','rb')) 
 96    pkg = Packager.ModelPackage(descCalc=calc,model=model) 
 97    pkg.SetNotes('General purpose model built from PhysProp data') 
 98    testD = [ 
 99      ('Fc1ccc(NC(=O)c2cccnc2Oc3cccc(c3)C(F)(F)F)c(F)c1',0,1.0 ), 
100      (r'CN/1(=C\C=C(/C=C1)\C\2=C\C=N(C)(Cl)\C=C2)Cl',0,0.70), 
101      (r'NS(=O)(=O)c1cc(ccc1Cl)C2(O)NC(=O)c3ccccc32',1,0.70), 
102      ] 
103       
104    for smi,pred,conf in testD: 
105      m = MolFromSmiles(smi) 
106      p,c = pkg.Classify(m) 
107      if pred!=p or conf!=c: 
108        raise ValueError,'Bad Prediction: %s'%(repr((smi,pred,conf,p,c))) 
109    cPickle.dump(pkg,open('test_data/Jan9_build3_pkg.pkl','wb+')) 
110    from Numeric import * 
111    import RandomArray 
112     
113    names = calc.GetDescriptorNames() 
114    perm = [names[x] for x in RandomArray.permutation(len(names))] 
115    calc.simpleList = perm 
116    calc.descriptorNames = perm 
117    pkg.Init() 
118    for smi,pred,conf in testD: 
119      m = MolFromSmiles(smi) 
120      p,c = pkg.Classify(m) 
121      if pred!=p or conf!=c: 
122        raise ValueError,'Bad Prediction: %s'%(repr((smi,pred,conf,p,c))) 
123