pcluster.py

   1 #!/usr/bin/env python
   2
   3 from mdp import pca
   4 from libhooke import WX_GOOD, ClickedPoint
   5 import wxversion
   6 wxversion.select(WX_GOOD)
   7 from wx import PostEvent
   8 import numpy as np
   9 import scipy as sp
  10 import copy
  11 import os.path
  12 import time
  13 import libhookecurve as lhc
  14
  15 import warnings
  16 warnings.simplefilter('ignore',np.RankWarning)
  17
  18
  19 class pclusterCommands:
  20
  21     def _plug_init(self):
  22         self.pca_paths=[]
  23         self.pca_myArray=None
  24         self.clustplot=None
  25
  26     def do_pcluster(self,args):
  27         '''
  28         pCLUSTER
  29         (pcluster.py)
  30         Automatically measures peaks and extracts informations for further clustering
  31         (c)Paolo Pancaldi, Massimo Sandal 2009
  32         '''
  33         #--Custom persistent length
  34         pl_value=None
  35         for arg in args.split():
  36             #look for a persistent length argument.
  37             if 'pl=' in arg:
  38                 pl_expression=arg.split('=')
  39                 pl_value=float(pl_expression[1]) #actual value
  40             else:
  41                 pl_value=None
  42
  43         #configuration variables
  44         min_npks = self.convfilt_config['minpeaks']
  45         min_deviation = self.convfilt_config['mindeviation']
  46
  47         pclust_filename=raw_input('Automeasure filename? ')
  48         realclust_filename=raw_input('Coordinates filename? ')
  49
  50         f=open(pclust_filename,'w+')
  51         f.write('Analysis started '+time.asctime()+'\n')
  52         f.write('----------------------------------------\n')
  53         f.write('; Contour length (nm)  ;  Persistence length (nm) ;  Max.Force (pN)  ;  Slope (N/m) ;  Sigma contour (nm) ; Sigma persistence (nm)\n')
  54         f.close()
  55
  56         f=open(realclust_filename,'w+')
  57         f.write('Analysis started '+time.asctime()+'\n')
  58         f.write('----------------------------------------\n')
  59         f.write('; Peak number ; Mean delta (nm)  ;  Median delta (nm) ;  Mean force (pN)  ;  Median force (pN) ; First peak length (nm) ; Last peak length (nm) ; Max force (pN) ; Min force (pN) ; Max delta (nm) ; Min delta (nm)\n')
  60         f.close()
  61         # ------ FUNCTION ------
  62         def fit_interval_nm(start_index,plot,nm,backwards):
  63             '''
  64             Calculates the number of points to fit, given a fit interval in nm
  65             start_index: index of point
  66             plot: plot to use
  67             backwards: if true, finds a point backwards.
  68             '''
  69             whatset=1 #FIXME: should be decidable
  70             x_vect=plot.vectors[1][0]
  71
  72             c=0
  73             i=start_index
  74             start=x_vect[start_index]
  75             maxlen=len(x_vect)
  76             while abs(x_vect[i]-x_vect[start_index])*(10**9) < nm:
  77                 if i==0 or i==maxlen-1: #we reached boundaries of vector!
  78                     return c
  79                 if backwards:
  80                     i-=1
  81                 else:
  82                     i+=1
  83                 c+=1
  84             return c
  85
  86         def plot_informations(itplot,pl_value):
  87             '''
  88             OUR VARIABLES
  89             contact_point.absolute_coords               (2.4584142802103689e-007, -6.9647135616234017e-009)
  90             peak_point.absolute_coords                  (3.6047748250571423e-008, -7.7142802788854212e-009)
  91             other_fit_point.absolute_coords     (4.1666139243838867e-008, -7.3759393477579707e-009)
  92             peak_location                                                                               [510, 610, 703, 810, 915, 1103]
  93             peak_size                                                                                           [-1.2729111505202212e-009, -9.1632775347399312e-010, -8.1707438353929907e-010, -8.0335812578148904e-010, -8.7483955226387558e-010, -3.6269619757067322e-009]
  94             params                                                                                                      [2.2433999931959462e-007, 3.3230248825175678e-010]
  95             fit_errors                                                                                  [6.5817195369767644e-010, 2.4415923138871498e-011]
  96             '''
  97             fit_points=int(self.config['auto_fit_points']) # number of points to fit before the peak maximum <50>
  98
  99             T=self.config['temperature'] #temperature of the system in kelvins. By default it is 293 K. <301.0>
 100             cindex=self.find_contact_point(itplot) #Automatically find contact point <158, libhooke.ClickedPoint>
 101             contact_point=self._clickize(itplot[0].vectors[1][0], itplot[0].vectors[1][1], cindex)
 102             self.basepoints=[]
 103             base_index_0=peak_location[-1]+fit_interval_nm(peak_location[-1], itplot[0], self.config['auto_right_baseline'],False)
 104             self.basepoints.append(self._clickize(itplot[0].vectors[1][0],itplot[0].vectors[1][1],base_index_0))
 105             base_index_1=self.basepoints[0].index+fit_interval_nm(self.basepoints[0].index, itplot[0], self.config['auto_left_baseline'],False)
 106             self.basepoints.append(self._clickize(itplot[0].vectors[1][0],itplot[0].vectors[1][1],base_index_1))
 107             self.basecurrent=self.current.path
 108             boundaries=[self.basepoints[0].index, self.basepoints[1].index]
 109             boundaries.sort()
 110             to_average=itplot[0].vectors[1][1][boundaries[0]:boundaries[1]] #y points to average
 111             avg=np.mean(to_average)
 112             return fit_points, contact_point, pl_value, T, cindex, avg
 113
 114         def features_peaks(itplot, peak, fit_points, contact_point, pl_value, T, cindex, avg):
 115             '''
 116             calculate informations for each peak and add they in
 117             c_lengths, p_lengths, sigma_c_lengths, sigma_p_lengths, forces, slopes
 118             '''
 119             c_leng=None
 120             p_leng=None
 121             sigma_c_leng=None
 122             sigma_p_leng=None
 123             force=None
 124             slope=None
 125
 126             delta_force=10
 127             slope_span=int(self.config['auto_slope_span'])
 128
 129             peak_point=self._clickize(itplot[0].vectors[1][0],itplot[0].vectors[1][1],peak)
 130             other_fit_point=self._clickize(itplot[0].vectors[1][0],itplot[0].vectors[1][1],peak-fit_points)
 131
 132             points=[contact_point, peak_point, other_fit_point]
 133
 134             params, yfit, xfit, fit_errors = self.wlc_fit(points, itplot[0].vectors[1][0], itplot[0].vectors[1][1], pl_value, T, return_errors=True)
 135
 136             #Measure forces
 137             delta_to_measure=itplot[0].vectors[1][1][peak-delta_force:peak+delta_force]
 138             y=min(delta_to_measure)
 139             #Measure slopes
 140             slope=self.linefit_between(peak-slope_span,peak)[0]
 141             #check fitted data and, if right, add peak to the measurement
 142             if len(params)==1: #if we did choose 1-value fit
 143                 p_leng=pl_value
 144                 c_leng=params[0]*(1.0e+9)
 145                 sigma_p_lengths=0
 146                 sigma_c_lengths=fit_errors[0]*(1.0e+9)
 147                 force = abs(y-avg)*(1.0e+12)
 148             else: #2-value fit
 149                 p_leng=params[1]*(1.0e+9)
 150                 #check if persistent length makes sense. otherwise, discard peak.
 151                 if p_leng>self.config['auto_min_p'] and p_leng<self.config['auto_max_p']:
 152                     '''
 153                     p_lengths.append(p_leng)
 154                     c_lengths.append(params[0]*(1.0e+9))
 155                     sigma_c_lengths.append(fit_errors[0]*(1.0e+9))
 156                     sigma_p_lengths.append(fit_errors[1]*(1.0e+9))
 157                     forces.append(abs(y-avg)*(1.0e+12))
 158                     slopes.append(slope)
 159                     '''
 160                     c_leng=params[0]*(1.0e+9)
 161                     sigma_c_leng=fit_errors[0]*(1.0e+9)
 162                     sigma_p_leng=fit_errors[1]*(1.0e+9)
 163                     force=abs(y-avg)*(1.0e+12)
 164                 else:
 165                     p_leng=None
 166                     slope=None
 167             #return c_lengths, p_lengths, sigma_c_lengths, sigma_p_lengths, forces, slopes
 168             return  c_leng, p_leng, sigma_c_leng, sigma_p_leng, force, slope
 169
 170
 171         # ------ PROGRAM -------
 172         c=0
 173         for item in self.current_list:
 174             c+=1
 175             item.identify(self.drivers)
 176             itplot=item.curve.default_plots()
 177             try:
 178                 peak_location,peak_size=self.exec_has_peaks(item,min_deviation)
 179             except:
 180                 #We have troubles with exec_has_peaks (bad curve, whatever).
 181                 #Print info and go to next cycle.
 182                 print 'Cannot process ',item.path
 183                 continue
 184
 185             if len(peak_location)==0:
 186                 print 'No peaks!'
 187                 continue
 188
 189             fit_points, contact_point, pl_value, T, cindex, avg = plot_informations(itplot,pl_value)
 190
 191             print '\n\nCurve',item.path, 'is',c,'of',len(self.current_list),': found '+str(len(peak_location))+' peaks.'
 192
 193             #initialize output data vectors
 194             c_lengths=[]
 195             p_lengths=[]
 196             sigma_c_lengths=[]
 197             sigma_p_lengths=[]
 198             forces=[]
 199             slopes=[]
 200
 201             #loop each peak of my curve
 202             for peak in peak_location:
 203                 c_leng, p_leng, sigma_c_leng, sigma_p_leng, force, slope = features_peaks(itplot, peak, fit_points, contact_point, pl_value, T, cindex, avg)
 204                 for var, vector in zip([c_leng, p_leng, sigma_c_leng, sigma_p_leng, force, slope],[c_lengths, p_lengths, sigma_c_lengths, sigma_p_lengths, forces, slopes]):
 205                     if var is not None:
 206                         vector.append(var)
 207
 208             #FIXME: We need a dictionary here...
 209             allvects=[c_lengths, p_lengths, sigma_c_lengths, sigma_p_lengths, forces, slopes]
 210             for vect in allvects:
 211                 if len(vect)==0:
 212                     for i in range(len(c_lengths)):
 213                         vect.append(0)
 214
 215             print 'Measurements for all peaks detected:'
 216             print 'contour (nm)', c_lengths
 217             print 'sigma contour (nm)',sigma_c_lengths
 218             print 'p (nm)',p_lengths
 219             print 'sigma p (nm)',sigma_p_lengths
 220             print 'forces (pN)',forces
 221             print 'slopes (N/m)',slopes
 222
 223             '''
 224             write automeasure text file
 225             '''
 226             print 'Saving automatic measurement...'
 227             f=open(pclust_filename,'a+')
 228             f.write(item.path+'\n')
 229             for i in range(len(c_lengths)):
 230                 f.write(' ; '+str(c_lengths[i])+' ; '+str(p_lengths[i])+' ; '+str(forces[i])+' ; '+str(slopes[i])+' ; '+str(sigma_c_lengths[i])+' ; '+str(sigma_p_lengths[i])+'\n')
 231             f.close()
 232
 233             '''
 234             calculate clustering coordinates
 235             '''
 236             peak_number=len(c_lengths)
 237             if peak_number > 1:
 238                 deltas=[]
 239                 for i in range(len(c_lengths)-1):
 240                     deltas.append(c_lengths[i+1]-c_lengths[i])
 241
 242                 delta_mean=np.mean(deltas)
 243                 delta_median=np.median(deltas)
 244
 245                 force_mean=np.mean(forces)
 246                 force_median=np.median(forces)
 247
 248                 first_peak_cl=c_lengths[0]
 249                 last_peak_cl=c_lengths[-1]
 250
 251                 max_force=max(forces[:-1])
 252                 min_force=min(forces)
 253
 254                 max_delta=max(deltas)
 255                 min_delta=min(deltas)
 256
 257                 delta_stdev=np.std(deltas)
 258                 forces_stdev=np.std(forces[:-1])
 259
 260                 print 'Coordinates'
 261                 print 'Peaks',peak_number
 262                 print 'Mean delta',delta_mean
 263                 print 'Median delta',delta_median
 264                 print 'Mean force',force_mean
 265                 print 'Median force',force_median
 266                 print 'First peak',first_peak_cl
 267                 print 'Last peak',last_peak_cl
 268                 print 'Max force',max_force
 269                 print 'Min force',min_force
 270                 print 'Max delta',max_delta
 271                 print 'Min delta',min_delta
 272                 print 'Delta stdev',delta_stdev
 273                 print 'Forces stdev',forces_stdev
 274
 275                 '''
 276                 write clustering coordinates
 277                 '''
 278                 f=open(realclust_filename,'a+')
 279                 f.write(item.path+'\n')
 280                 f.write(' ; '+str(peak_number)+' ; '+str(delta_mean)+' ; '+str(delta_median)+' ; '+str(force_mean)+' ; '+str(force_median)+' ; '+str(first_peak_cl)+' ; '+str(last_peak_cl)+ ' ; '+str(max_force)+' ; '
 281                 +str(min_force)+' ; '+str(max_delta)+' ; '+str(min_delta)+ ' ; '+str(delta_stdev)+ ' ; '+str(forces_stdev)+'\n')
 282                 f.close()
 283             else:
 284                 pass
 285
 286     def do_pca(self,args):
 287         '''
 288         PCA
 289         Automatically measures pca from coordinates filename and shows two interactives plots
 290         (c)Paolo Pancaldi, Massimo Sandal 2009
 291         '''
 292
 293         self.pca_myArray = []
 294         self.pca_paths = {}
 295         plot_path_temp = ""
 296         nPlotTot = 0
 297         nPlotGood = 0
 298
 299         file_name=args
 300
 301         for arg in args.split():
 302             #look for a file argument.
 303             if 'f=' in arg:
 304                 file_temp=arg.split('=')[1] #actual coordinates filename
 305                 try:
 306                     f=open(file_temp)
 307                     f.close()
 308                     file_name = file_temp
 309                     print "Coordinates filename used: " + file_name
 310                 except:
 311                     print "Impossibile to find " + file_temp + " in current directory"
 312                     print "Coordinates filename used: " + file_name
 313
 314         f=open(file_name)
 315         rows = f.readlines()
 316         for row in rows:
 317             if row[0]=="/":
 318                 nPlotTot = nPlotTot+1
 319                 #plot_path_temp = row.split("/")[6][:-1]
 320                 plot_path_temp = row
 321             if row[0]==" " and row.find('nan')==-1:
 322                 row = row[row.index(";",2)+2:].split(" ; ")     # non considero la prima colonna col #picchi
 323                 row = [float(i) for i in row]
 324
 325                 #0:Mean delta, 1:Median delta, 2:Mean force, 3:Median force, 4:First peak length, 5:Last peak length
 326                         #6:Max delta 7:Min delta 8:Max force 9:Min force 10:Std delta 11:Std force
 327                 if (row[0]<9000 and row[1]<9000 and row[2]<9000 and row[3]<9000 and row[4]<9000 and row[5]<9000):
 328                     if (row[0]>0 and row[1]>0 and row[2]>0 and row[3]>0 and row[4]>0 and row[5]>0):
 329                         self.pca_paths[nPlotGood] = plot_path_temp
 330                         row=[row[1],row[2]]
 331                         self.pca_myArray.append(row)
 332                         nPlotGood = nPlotGood+1
 333
 334         f.close()
 335         print nPlotGood, "of", nPlotTot
 336
 337         # array convert, calculate PCA, transpose
 338         self.pca_myArray = np.array(self.pca_myArray,dtype='float')
 339         print self.pca_myArray.shape
 340         '''for i in range(len(self.pca_myArray)):
 341             print i, self.pca_paths[i]
 342             print i, self.pca_myArray[i]'''
 343         self.pca_myArray = pca(self.pca_myArray, output_dim=2)  #other way -> y = mdp.nodes.PCANode(output_dim=2)(gigi)
 344         myArrayTr = np.transpose(self.pca_myArray)
 345
 346         '''for i in range(len(self.pca_myArray)):
 347             print i, self.pca_paths[i]
 348             print i, self.pca_myArray[i]'''
 349
 350         # plotting
 351         X=myArrayTr[0]
 352         Y=myArrayTr[1]
 353         clustplot=lhc.PlotObject()
 354
 355         #FIXME
 356         #our dataset-specific stuff
 357         #This will go away after testing :)
 358         Xsyn=[]
 359         Ysyn=[]
 360         Xgb1=[]
 361         Ygb1=[]
 362         for index in range(len(self.pca_paths)):
 363             if 'syn' in self.pca_paths[index]:
 364                 Xsyn.append(X[index])
 365                 Ysyn.append(Y[index])
 366             else:
 367                 Xgb1.append(X[index])
 368                 Ygb1.append(Y[index])
 369
 370         clustplot.add_set(Xsyn,Ysyn)
 371         clustplot.add_set(Xgb1,Ygb1)
 372         clustplot.normalize_vectors()
 373         clustplot.styles=['scatter', 'scatter']
 374         clustplot.colors=[None,'red']
 375         #clustplot.styles=['scatter',None]
 376         clustplot.destination=1
 377         self._send_plot([clustplot])
 378         self.clustplot=clustplot
 379
 380
 381     def do_pclick(self,args):
 382
 383         self._send_plot([self.clustplot]) #quick workaround for BAD problems in the GUI
 384         print 'Click point'
 385         point = self._measure_N_points(N=1, whatset=0)
 386         indice = point[0].index
 387         plot_file = self.pca_paths[indice]
 388         dot_coord = self.pca_myArray[indice]
 389         print "file: " + str(plot_file).rstrip()
 390         print "id: " + str(indice)
 391         print "coord: " + str(dot_coord)
 392         self.do_genlist(str(plot_file))
 393         #self.do_jump(str(plot_file))