def add(self, E, selected=None, overlay=False): """Add ekin proj to plot""" total = len(E.datasets) if total == 0: return if selected == None: datasets = E.datasets else: datasets = self.getNames(selected, E) if len(datasets) > 25: print 'too many plots' datasets = datasets[:25] elif len(E.datasets) == 1: datasets = E.datasets[0] if len(datasets) == 0: datasets = E.datasets if overlay == True: plotopt = 3 self.Opt.opts['legend'] = True else: plotopt = 2 fr = Frame(self.pw) self.pw.add(fr) plotframe = PlotPanel(parent=fr, side=BOTTOM, tools=True) plotframe.setProject(E) plotframe.Opts.opts = self.Opt.opts plotframe.Opts.opts['title'] = E.name plotframe.Opts.opts['normalise'] = overlay plotframe.plotCurrent(datasets=datasets, plotoption=plotopt) self.plotframes[E.name] = plotframe return plotframe
def add(self, E, selected=None, overlay=False): """Add ekin proj to plot""" total = len(E.datasets) if total == 0: return if selected == None: datasets = E.datasets else: datasets = self.getNames(selected, E) if len(datasets)>25: print 'too many plots' datasets = datasets[:25] elif len(E.datasets)==1: datasets = E.datasets[0] if len(datasets) == 0: datasets = E.datasets if overlay == True: plotopt = 3 self.Opt.opts['legend'] = True else: plotopt = 2 fr=Frame(self.pw) self.pw.add(fr) plotframe = PlotPanel(parent=fr, side=BOTTOM, tools=True) plotframe.setProject(E) plotframe.Opts.opts = self.Opt.opts plotframe.Opts.opts['title'] = E.name plotframe.Opts.opts['normalise']=overlay plotframe.plotCurrent(datasets=datasets, plotoption=plotopt) self.plotframes[E.name] = plotframe return plotframe
class VantHoff(Plugin): """A plugin to do Van't Hoff Analysis of temperature melting curves""" """Author: Damien Farrell""" capabilities = ['gui','uses_sidepane'] requires = ['pylab','numpy'] menuentry = "Van't Hoff Analysis" gui_methods = {'getCSV': 'Import CSV', 'loadEkin':'Load Ekin Proj', 'saveEkin':'Save Ekin Proj', 'doAnalysis':"Do Analysis", #'benchmark': 'Do Benchmark', 'close':'Close' } about = "A plugin to do Van't Hoff Analysis of temperature melting curves" R = 8.3144 def __init__(self): self.path = os.path.expanduser("~") self.pltConfig() self.E = None return def main(self, parent): if parent==None: return self.parent = parent self.DB = parent.DB self.xydata = None self._doFrame() return def _doFrame(self): if 'uses_sidepane' in self.capabilities: self.mainwin = self.parent.createChildFrame(width=600) else: self.mainwin=Toplevel() self.mainwin.title(self.menuentry) self.mainwin.geometry('800x600+200+100') methods = self._getmethods() fr = Frame(self.mainwin) fr.pack(side=LEFT,fill=BOTH) methods = [m for m in methods if m[0] in self.gui_methods.keys()] self._createButtons(methods, fr) self.showDatasetSelector() self.doall = Pmw.RadioSelect(fr, buttontype = 'checkbutton', orient = 'horizontal', labelpos = 'w') self.doall.add('Process All') self.doall.pack() self.conversions = Pmw.RadioSelect(fr, buttontype = 'checkbutton', orient = 'horizontal', labelpos = 'w') self.conversions.add('Convert Celsius-Kelvin') self.conversions.pack() self.methods = Pmw.RadioSelect(fr, buttontype = 'checkbutton', orient = 'vertical', labelpos = 'w', label_text = 'Methods:') for m in ['method 1','method 2','method 3', 'method 4']: self.methods.add(m) self.methods.invoke('method 1') self.methods.pack() self.sm = Pmw.EntryField(fr, labelpos = 'w', value = 5, label_text = 'Smoothing:') self.sm.pack() self.tw = Pmw.EntryField(fr, labelpos = 'w', value = 60, label_text = 'Width of transition:') self.tw.pack() return def _getmethods(self): """Get a list of all available public methods""" import inspect mems = inspect.getmembers(self, inspect.ismethod) methods = [m for m in mems if not m[0].startswith('_')] return methods def _createButtons(self, methods, fr=None): """Dynamically create buttons for supplied methods, which is a tuple of (method name, label)""" for m in methods: b=Button(fr,text=self.gui_methods[m[0]],command=m[1]) b.pack(side=TOP,fill=BOTH) return def close(self): self.mainwin.destroy() self.plotframe = None return def showDatasetSelector(self): if self.E==None: return if hasattr(self, 'dmenu'): self.dmenu.destroy() self.dmenu = Pmw.OptionMenu(self.mainwin, labelpos = 'w', label_text = 'Dataset:', items = sorted(self.E.datasets), command=self.showPreview, menubutton_width = 8) self.dmenu.pack(side=TOP,fill=BOTH) return def showPreview(self,event=None): if self.E == None: return if not hasattr(self, 'plotframe') or self.plotframe == None: from Ekin.Plotting import PlotPanel self.plotframe = PlotPanel(parent=self.mainwin, side=BOTTOM) self.plotframe.setProject(self.E) d = self.dmenu.getcurselection() self.plotframe.plotCurrent(d) #plt.close(1) return def getCSV(self): """Import a csv file""" self.E = EkinProject() from PEATDB.Ekin.IO import Importer importer = Importer(self,parent_win=self.mainwin) newdata = importer.import_multiple() if newdata == None: return for n in newdata.keys(): self.E.insertDataset(newdata[n], n, update=None) print 'imported %s datasets' %len(self.E.datasets) self.showDatasetSelector() self.showPreview() return def loadEkin(self): """Load the ekin prj""" filename=tkFileDialog.askopenfilename(defaultextension='.ekinprj', initialdir=os.getcwd(), filetypes=[("ekinprj","*.ekinprj"), ("All files","*.*")], parent=self.mainwin) if not os.path.isfile(filename): return self.E = EkinProject() self.E.openProject(filename) self.showDatasetSelector() self.showPreview() return def saveEkin(self): """save proj""" if self.E != None: if self.E.filename == None: self.E.filename = tkFileDialog.asksaveasfilename(defaultextension='.ekinprj', initialdir=os.getcwd(), filetypes=[("ekinprj","*.ekinprj"), ("All files","*.*")], parent=self.mainwin) self.E.saveProject() print 'saved ekin proj' return def doAnalysis(self): """Execute from GUI""" if self.E == None: return methods = self.methods.getcurselection() if 'Process All' in self.doall.getcurselection(): self.doAll(methods=methods) else: if 'method 1' in methods: self.fitVantHoff(E=self.E,d=self.dmenu.getcurselection(), transwidth=int(self.tw.getvalue())) if 'method 2' in methods: self.fitElwellSchellman(E=self.E,d=self.dmenu.getcurselection(), transwidth=int(self.tw.getvalue())) if 'method 3' in methods: self.fitDifferentialCurve(E=self.E,d=self.dmenu.getcurselection(), smooth=int(self.sm.getvalue())) if 'method 4' in methods: self.breslauerMethod(E=self.E,d=self.dmenu.getcurselection())#,invert=opts.invert) return def guessMidpoint(self,x,y): """guess midpoint for unfolding model""" midy=min(y)+(max(y)-min(y))/2.0 midx=0 closest=1e4 for i in range(len(x)): c=abs(y[i]-midy) if c<closest: midx=x[i] closest=c return midx def transformCD(self,x,y,transwidth=None,ax=None): """Transform raw data into fraction unfolded per temp value, by fitting to a general unfolding equation that extracts baseline/slopes""" #fit baseline slopes and get intercepts d50 = self.guessMidpoint(x,y) print 'fitting to get baseline slopes and intercepts..' print 'midpoint is %s' %d50 A,X=Fitting.doFit(expdata=zip(x,y),model='Unfolding',noiter=50,silent=True, guess=False,startvalues=[1,1,1,1,1,d50]) #print X.getResult() fity = X.getFitLine(x) fd=X.getFitDict() if ax!=None: p=ax.plot(x,fity,'r',lw=2) self.drawParams(ax,fd) #we then use slopes and intercepts get frac unfolded at each temp mn = fd['bn']; mu = fd['bd'] #slopes #if mu>0.01: mu = 0.01 yn = fd['an']; yu = fd['ad'] #intercepts d50 = fd['d50']; m = fd['m'] t=[]; f=[] #print mu, mn for T,yo in zip(x,y): fu = (yo-(yn+mn*T)) / ((yu+mu*T)-(yn+mn*T)) #print fu, (yo-(yn+mn*T)), (m), mu, mn #if f>0: f.append(fu) t.append(T) #try to take useful transition region of data at,af=t,f diff=1e5 if transwidth != None: for i in t: d=abs(i-d50) if d<diff: mid = t.index(i) diff=d L=int(mid-transwidth); U=int(mid+transwidth) t,f = t[L:U], f[L:U] return at,af,t,f def fitVantHoff(self, E=None, d=None, xy=None, transwidth=80, invert=False, show=True, figname=None): """Derive fraction unfolded, get K and fit to Van't Hoff. see http://www.jbc.org/content/277/43/40717.full or http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2144003/ """ if E != None: if not d in E.datasets: print 'no such dataset, %s' %d print 'available datasets:', E.datasets return ek = E.getDataset(d) x,y = ek.getxySorted() elif xy!=None: x,y = xy if 'Convert Celsius-Kelvin' in self.conversions.getcurselection(): x = [i+273 for i in x] if invert == True: y = [max(y)-i for i in y[:]] f=plt.figure(figsize=(18,6)) ax=f.add_subplot(131) p=ax.plot(x,y,'o',alpha=0.6) ax.set_xlabel('T(K)'); ax.set_ylabel('mdeg') ax.set_title('raw data') x1,y1,x,y = self.transformCD(x,y,transwidth,ax) cw=csv.writer(open('frac_unfolded_'+d+'.csv','w')) cw.writerow(['temp','frac']) for i in zip(x1,y1): cw.writerow(i) #derive lnK vs 1/T t=[]; k=[] for T,fu in zip(x,y): if fu>=1 or fu<=0: continue K = fu/(1-fu) klog = math.log(K) k.append(klog) t.append(1/T) if len(t)<2: return None, None, None ax=f.add_subplot(132) p=ax.plot(x1,y1,'o',color='g',alpha=0.6) ax.set_xlabel('T(K)'); ax.set_ylabel('fu') ax.set_title('fraction unfolded') ax=f.add_subplot(133) p=ax.plot(t,k,'x',mew=2,color='black') ax.set_xlabel('1/T')#(r'$1/T ($K^-1)$') ax.set_ylabel('ln K') formatter = matplotlib.ticker.ScalarFormatter() formatter.set_scientific(True) formatter.set_powerlimits((0,0)) ax.xaxis.set_major_formatter(formatter) for l in ax.get_xticklabels(): l.set_rotation(30) #fit this van't hoff plot A,X=Fitting.doFit(expdata=zip(t,k),model='Linear') fitk = X.getFitLine(t) p=ax.plot(t,fitk,'r',lw=2) fd=X.getFitDict() #self.drawParams(ax,fd) #slope is deltaH/R/1000 in kJ deltaH = -fd['a']*self.R/1000 deltaS = fd['b']*self.R/1000 f.suptitle("Method 1 - deltaH: %2.2f deltaS: %2.2f" %(deltaH,deltaS),size=18) f.subplots_adjust(bottom=0.15,top=0.85) if show==True: self.showTkFigure(f) if figname == None: figname = d figname = figname.replace('.','_') fname = figname+'m1'+'.png' f.savefig(fname,dpi=300) print 'plot saved to %s' %os.path.abspath(fname) #plt.close() if E!=None: fdata = Fitting.makeFitData(X.name,vrs=X.variables) E.insertDataset(xydata=[t,k], newname=d+'_vanthoff',replace=True,fit=fdata) #E.saveProject() return deltaH, deltaS, ax def fitElwellSchellman(self,E=None, d=None, xy=None,transwidth=50, invert=False,show=True,figname=None): """Fit entire raw data simultaneously to the three main thermodynamic parameters using Elwell/Schellman method""" if E !=None: ek = E.getDataset(d) x,y,a, xerr,yerr = ek.getAll() elif xy!=None: x,y = xy else: return if invert == True: y = [max(y)-i for i in y[:]] f=plt.figure(figsize=(10,5)) ax=f.add_subplot(121) p=ax.plot(x,y,'o',alpha=0.5) ax.set_xlabel('T');ax.set_xlabel('mdeg') ax.set_title('raw data') x1,y1,x,y = self.transformCD(x,y,transwidth,ax) t=[];dg=[] R=8.3144e-3 for T,fu in zip(x,y): if fu>=1 or fu<=0: continue K = fu/(1-fu) deltaGt = -R * T * math.log(K) dg.append(deltaGt) t.append(T) ax1=f.add_subplot(122) p=ax1.plot(t,dg,'x',mew=2,color='black') ax1.set_xlabel('T'); ax1.set_ylabel('dG(T)') ax.set_title('stability curve') A,X=Fitting.doFit(expdata=zip(t,dg),model='schellman',grad=1e-9,conv=1e-9) fity = X.getFitLine(t) p=ax1.plot(t,fity,'r',lw=2) fd=X.getFitDict() self.drawParams(ax1,fd) deltaH=fd['deltaH']; deltacp=fd['deltacp']; Tm=fd['Tm'] f.suptitle("Method 2 - deltaH: %2.2f deltaCp: %2.2e Tm: %2.2f" %(deltaH,deltacp,Tm),size=18) if show == True: self.showTkFigure(f) if figname == None: figname = d figname = figname.replace('.','_') fname = figname+'m1'+'.png' f.savefig(fname,dpi=300) print 'plot saved to %s' %os.path.abspath(fname) if E!=None: fdata = Fitting.makeFitData(X.name,vrs=X.variables) E.insertDataset(xydata=[t,dg], newname=d+'_vanthoff2',replace=True,fit=fdata) #E.saveProject() return deltaH, Tm, deltacp def breslauerMethod(self,E=None, d=None, xy=None,invert=False, show=True,figname=None): """Finds slope of trans region and plugs this in to equation http://www.springerlink.com/content/r34n0201g30563u7/ """ if E !=None: ek = E.getDataset(d) x,y,a, xerr,yerr = ek.getAll() elif xy!=None: x,y = xy else: return f=plt.figure(figsize=(10,6)) ax=f.add_subplot(111) ax.set_xlabel('T') p=ax.plot(x,y,'o',alpha=0.5) d50 = self.guessMidpoint(x,y) A,X=Fitting.doFit(expdata=zip(x,y),model='Unfolding',conv=1e-7,noiter=60, guess=False,startvalues=[1,1,1,1,1,d50]) fity = X.getFitLine(x) p=ax.plot(x,fity,'r',lw=2) fd=X.getFitDict() self.drawParams(ax,fd) Tm = fd['d50']; m = fd['m'] R = 8.3144e-3 deltaH = R * math.pow(Tm,2) * m f.suptitle("Method 4 - deltaH: %2.2f Tm: %2.2f" %(deltaH,Tm),size=18) if show == True: self.showTkFigure(f) if figname != None: figname = figname.replace('.','_') f.savefig(figname) plt.close() return deltaH, Tm def fitDifferentialCurve(self, E=None, d=None, xy=None,smooth=0, invert=False,show=True,figname=None): """Derive differential denaturation curve and fit to get deltaH We smooth the unfolding curve and then differentiate and finally fit to a 3 parameter equation. See http://www.ncbi.nlm.nih.gov/pubmed/10933511""" if E !=None: ek = E.getDataset(d) x,y,a, xerr,yerr = ek.getAll() elif xy!=None: x,y = xy else: return if invert == True: y = [max(y)-i for i in y[:]] leg=[]; lines=[] f=plt.figure(figsize=(10,5)) ax=f.add_subplot(121) p=ax.plot(x,y,'x',color='black',mew=3,alpha=0.5) leg.append(p); lines.append('original') #smooth if smooth == 0: smooth=int(len(x)/15.0) s=self.smoothListGaussian(y,smooth) p=ax.plot(x[:len(s)-1],s[:-1],lw=3) leg.append(p); lines.append('smoothed') ax.set_title("original data") ax.set_xlabel('T') ax1=f.add_subplot(122) #differentiate dx,ds = self.differentiate(x[:len(s)],s) #ds = [i/max(ds) for i in ds] ds = [i*10 for i in ds] cw=csv.writer(open('diffcd.csv','w')) for row in zip(dx,ds): cw.writerow(row) p=ax1.plot(dx,ds,'-',lw=1.5,alpha=0.7,color='black') leg.append(p); lines.append('differential') ax1.set_title("differential denaturation") ax1.set_xlabel('T'); ax1.set_ylabel('dsignal/dT') A,X=Fitting.doFit(expdata=zip(dx,ds),model='diffDenaturation',grad=1e-9,conv=1e-10) fity = X.getFitLine(dx) p=ax1.plot(dx,fity,'r',lw=2) leg.append(p); lines.append('fit') t=X.getFitDict() self.drawParams(ax1,t) dHkcal=t['deltaH']/4.184 f.suptitle('Method 3 - deltaH: %2.2f kJ/mol (%2.2f kcal) Tm: %2.2f' %(t['deltaH'],dHkcal,t['Tm']),size=18) ax.legend(leg,lines,loc='best',prop=FontProperties(size="smaller")) #f.subplots_adjust(hspace=0.8) if show == True: self.showTkFigure(f) if figname != None: figname = figname.replace('.','_') f.savefig(figname+'m3',dpi=300) plt.close() if E!=None: fdata = Fitting.makeFitData(X.name,vrs=X.variables) E.insertDataset(xydata=[dx,ds], newname=d+'_diff',replace=True,fit=fdata) #E.saveProject() return t['deltaH'],t['Tm'] def differentiate(self, x,y): dy = numpy.diff(y,1) dx = x[:len(dy)] return dx,dy def smoothListGaussian(self,data,degree=5): """Gaussian data smoothing function""" #buffer data to avoid offset result data=list(data) data = [data[0]]*(degree-1) + data + [data[-1]]*degree window=degree*2-1 weight=numpy.array([1.0]*window) weightGauss=[] for i in range(window): i=i-degree+1 frac=i/float(window) gauss=1/(numpy.exp((4*(frac))**2)) weightGauss.append(gauss) weight=numpy.array(weightGauss)*weight smoothed=[0.0]*(len(data)-window) for i in range(len(smoothed)): smoothed[i]=sum(numpy.array(data[i:i+window])*weight)/sum(weight) return smoothed def invert(self,data): inv=[i for i in data] return inv def simulateCD(self,noise=1.0): """Simulate some CD spec data""" x=list(numpy.arange(290,380,0.2)); y=[] X=Fitting.getFitter(model='Unfolding', vrs=[-16, 0.01, -11.6, 0.01, 2.7, 324]) fity = X.getFitLine(x) for i in fity: noise=numpy.random.normal(i, 1.0/2) y.append(i+noise) cw=csv.writer(open('cd.csv','w')) for row in zip(x,y): cw.writerow(row) return x,y def drawParams(self,ax,d): ymin, ymax = ax.get_ylim() xmin, xmax = ax.get_xlim() inc=(ymax-ymin)/20 xinc=(xmax-xmin)/20 y=ymax-inc for k in d: s = k+'='+str(round(d[k],3)) ax.text(xmin+xinc,y,s,fontsize=10) y-=inc return def pltConfig(self): #plt.rc('text', usetex=True) plt.rc('figure.subplot', hspace=0.3,wspace=0.3) #plt.rc('axes',titlesize=22) plt.rc('font',family='monospace') return def doAll(self, methods=['method 1']): """Process all datasets in ekinprj""" E=self.E vals={} from Dialogs import PEATDialog pb=PEATDialog(self.mainwin, option='progressbar', message='Analysing Data..') pb.update_progress(0) total = len(E.datasets); count=0 for d in E.datasets: if '_diff' in d or '_vanthoff' in d: continue vals[d]={} name = d if 'method 1' in methods: vals[d]['dH1'], vals[d]['dS1'], ax = self.fitVantHoff(E,d, transwidth=int(self.tw.getvalue()), show=False,figname=name) if 'method 2' in methods: vals[d]['dH2'], vals[d]['dTm2'], vals[d]['dCp2'] = self.fitElwellSchellman(E,d,show=False,figname=name) if 'method 3' in methods: vals[d]['dH3'], vals[d]['dTm3'] = self.fitDifferentialCurve(E,d,show=False,figname=name) count += 1 pb.update_progress(float(count)/total*100.0) pb.close() self.showTable(vals) return def showTable(self, data): """Show results in table""" from PEATDB.DictEdit import DictEditor D=DictEditor(self.mainwin) D.loadTable(data) return def benchmark(self,E=None,d=None, method=1): """Test methods with varying paramaters, smoothing etc""" if E==None and self.E != None: E = self.E; d=self.dmenu.getcurselection() path='vh_benchmark' if not os.path.exists(path): os.mkdir(path) dHvals=[] if method == 1: xlabel = 'width (K)' title = 'method 1: deltaH variation with trans region width fit' vals=range(5,140,5) for w in vals: dH, dS, ax = self.fitVantHoff(E,d,transwidth=w,show=False, figname=os.path.join(path,'%s_%s.png' %(d,w))) if dH == None: dH=0 dHvals.append(dH) #take best values from middle #dHvals= dHvals[5:16] elif method == 2: xlabel = 'width (K)' title = 'method 2: deltaH variation with width fit' vals=range(5,140,5) for w in vals: dH, dcp, dTm = self.fitElwellSchellman(E,d,transwidth=w,show=False, figname=os.path.join(path,'%s_%s.png' %(d,w))) dHvals.append(dH) elif method == 3: xlabel = 'smoothing degree' title = 'method 3: deltaH variation with degree of smoothing' vals=range(1,30,3) for s in vals: dH, dTm = self.fitDifferentialCurve(E,d,smooth=s,show=False, figname=os.path.join(path,'%s_%s.png' %(d,s))) dHvals.append(dH) mean = numpy.mean(dHvals) stdev = numpy.std(dHvals) f=plt.figure() ax=f.add_subplot(111) ax.plot(vals, dHvals,lw=2) ax.set_xlabel(xlabel) ax.set_ylabel('deltaH (kJ)') ax.set_title('mean: %2.2f stdev: %2.2f'%(mean, stdev)) f.suptitle(title) f.savefig('benchmark_%s.png' %method) cw=csv.writer(open('benchmark_%s.csv' %method,'w')) for row in zip(vals,dHvals): cw.writerow(row) return def benchmarkLimitedData(self, E=None,d=None, method=1): """test any method with varying limited data""" if E==None and self.E != None: E = self.E; d=self.dmenu.getcurselection() path='vh_benchmark' if not os.path.exists(path): os.mkdir(path) dHvals=[] vals=[] if method == 1: L=range(5,140,5) for w in vals: dH, dS, ax = self.fitVantHoff(E,d,transwidth=w,show=False, figname=os.path.join(path,'%s_%s.png' %(d,w))) return @classmethod def plotCorrelation(self,x=None,y=None,xlabel='method1',ylabel='method2'): if x==None: data=open('compared.csv','r') cr=csv.reader(data) x=[float(r[0]) for r in cr]; data.seek(0) y=[float(r[1]) for r in cr] f=plt.figure() ax=f.add_subplot(111) line = ax.scatter(x, y, marker='o',alpha=0.8) cl = numpy.arange(0,max(x)+50) ax.plot(cl, cl, 'g', alpha=0.5,lw=2) ax.set_xlabel(xlabel) ax.set_ylabel(ylabel) ax.set_xlim(150,600); ax.set_ylim(150,600) ax.set_title('Correlation') from scipy.stats import stats cc = str(round(pow(stats.pearsonr(x,y)[0],2),2)) ax.text(400,180, r'$r^2= %s$' %cc, fontsize=16) self.showTkFigure(f) return def showTkFigure(self, fig): from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg, NavigationToolbar2TkAgg fr = Toplevel() canvas = FigureCanvasTkAgg(fig, master=fr) #self.canvas.show() canvas.get_tk_widget().pack(side=TOP, fill=X, expand=1) mtoolbar = NavigationToolbar2TkAgg(canvas, fr) mtoolbar.update() canvas._tkcanvas.pack(side=BOTTOM, fill=BOTH, expand=1) return
class VantHoff(Plugin): """A plugin to do Van't Hoff Analysis of temperature melting curves""" """Author: Damien Farrell""" capabilities = ['gui', 'uses_sidepane'] requires = ['pylab', 'numpy'] menuentry = "Van't Hoff Analysis" gui_methods = { 'getCSV': 'Import CSV', 'loadEkin': 'Load Ekin Proj', 'saveEkin': 'Save Ekin Proj', 'doAnalysis': "Do Analysis", #'benchmark': 'Do Benchmark', 'close': 'Close' } about = "A plugin to do Van't Hoff Analysis of temperature melting curves" R = 8.3144 def __init__(self): self.path = os.path.expanduser("~") self.pltConfig() self.E = None return def main(self, parent): if parent == None: return self.parent = parent self.DB = parent.DB self.xydata = None self._doFrame() return def _doFrame(self): if 'uses_sidepane' in self.capabilities: self.mainwin = self.parent.createChildFrame(width=600) else: self.mainwin = Toplevel() self.mainwin.title(self.menuentry) self.mainwin.geometry('800x600+200+100') methods = self._getmethods() fr = Frame(self.mainwin) fr.pack(side=LEFT, fill=BOTH) methods = [m for m in methods if m[0] in self.gui_methods.keys()] self._createButtons(methods, fr) self.showDatasetSelector() self.doall = Pmw.RadioSelect(fr, buttontype='checkbutton', orient='horizontal', labelpos='w') self.doall.add('Process All') self.doall.pack() self.conversions = Pmw.RadioSelect(fr, buttontype='checkbutton', orient='horizontal', labelpos='w') self.conversions.add('Convert Celsius-Kelvin') self.conversions.pack() self.methods = Pmw.RadioSelect(fr, buttontype='checkbutton', orient='vertical', labelpos='w', label_text='Methods:') for m in ['method 1', 'method 2', 'method 3', 'method 4']: self.methods.add(m) self.methods.invoke('method 1') self.methods.pack() self.sm = Pmw.EntryField(fr, labelpos='w', value=5, label_text='Smoothing:') self.sm.pack() self.tw = Pmw.EntryField(fr, labelpos='w', value=60, label_text='Width of transition:') self.tw.pack() return def _getmethods(self): """Get a list of all available public methods""" import inspect mems = inspect.getmembers(self, inspect.ismethod) methods = [m for m in mems if not m[0].startswith('_')] return methods def _createButtons(self, methods, fr=None): """Dynamically create buttons for supplied methods, which is a tuple of (method name, label)""" for m in methods: b = Button(fr, text=self.gui_methods[m[0]], command=m[1]) b.pack(side=TOP, fill=BOTH) return def close(self): self.mainwin.destroy() self.plotframe = None return def showDatasetSelector(self): if self.E == None: return if hasattr(self, 'dmenu'): self.dmenu.destroy() self.dmenu = Pmw.OptionMenu(self.mainwin, labelpos='w', label_text='Dataset:', items=sorted(self.E.datasets), command=self.showPreview, menubutton_width=8) self.dmenu.pack(side=TOP, fill=BOTH) return def showPreview(self, event=None): if self.E == None: return if not hasattr(self, 'plotframe') or self.plotframe == None: from Ekin.Plotting import PlotPanel self.plotframe = PlotPanel(parent=self.mainwin, side=BOTTOM) self.plotframe.setProject(self.E) d = self.dmenu.getcurselection() self.plotframe.plotCurrent(d) #plt.close(1) return def getCSV(self): """Import a csv file""" self.E = EkinProject() from PEATDB.Ekin.IO import Importer importer = Importer(self, parent_win=self.mainwin) newdata = importer.import_multiple() if newdata == None: return for n in newdata.keys(): self.E.insertDataset(newdata[n], n, update=None) print 'imported %s datasets' % len(self.E.datasets) self.showDatasetSelector() self.showPreview() return def loadEkin(self): """Load the ekin prj""" filename = tkFileDialog.askopenfilename(defaultextension='.ekinprj', initialdir=os.getcwd(), filetypes=[ ("ekinprj", "*.ekinprj"), ("All files", "*.*") ], parent=self.mainwin) if not os.path.isfile(filename): return self.E = EkinProject() self.E.openProject(filename) self.showDatasetSelector() self.showPreview() return def saveEkin(self): """save proj""" if self.E != None: if self.E.filename == None: self.E.filename = tkFileDialog.asksaveasfilename( defaultextension='.ekinprj', initialdir=os.getcwd(), filetypes=[("ekinprj", "*.ekinprj"), ("All files", "*.*")], parent=self.mainwin) self.E.saveProject() print 'saved ekin proj' return def doAnalysis(self): """Execute from GUI""" if self.E == None: return methods = self.methods.getcurselection() if 'Process All' in self.doall.getcurselection(): self.doAll(methods=methods) else: if 'method 1' in methods: self.fitVantHoff(E=self.E, d=self.dmenu.getcurselection(), transwidth=int(self.tw.getvalue())) if 'method 2' in methods: self.fitElwellSchellman(E=self.E, d=self.dmenu.getcurselection(), transwidth=int(self.tw.getvalue())) if 'method 3' in methods: self.fitDifferentialCurve(E=self.E, d=self.dmenu.getcurselection(), smooth=int(self.sm.getvalue())) if 'method 4' in methods: self.breslauerMethod( E=self.E, d=self.dmenu.getcurselection()) #,invert=opts.invert) return def guessMidpoint(self, x, y): """guess midpoint for unfolding model""" midy = min(y) + (max(y) - min(y)) / 2.0 midx = 0 closest = 1e4 for i in range(len(x)): c = abs(y[i] - midy) if c < closest: midx = x[i] closest = c return midx def transformCD(self, x, y, transwidth=None, ax=None): """Transform raw data into fraction unfolded per temp value, by fitting to a general unfolding equation that extracts baseline/slopes""" #fit baseline slopes and get intercepts d50 = self.guessMidpoint(x, y) print 'fitting to get baseline slopes and intercepts..' print 'midpoint is %s' % d50 A, X = Fitting.doFit(expdata=zip(x, y), model='Unfolding', noiter=50, silent=True, guess=False, startvalues=[1, 1, 1, 1, 1, d50]) #print X.getResult() fity = X.getFitLine(x) fd = X.getFitDict() if ax != None: p = ax.plot(x, fity, 'r', lw=2) self.drawParams(ax, fd) #we then use slopes and intercepts get frac unfolded at each temp mn = fd['bn'] mu = fd['bd'] #slopes #if mu>0.01: mu = 0.01 yn = fd['an'] yu = fd['ad'] #intercepts d50 = fd['d50'] m = fd['m'] t = [] f = [] #print mu, mn for T, yo in zip(x, y): fu = (yo - (yn + mn * T)) / ((yu + mu * T) - (yn + mn * T)) #print fu, (yo-(yn+mn*T)), (m), mu, mn #if f>0: f.append(fu) t.append(T) #try to take useful transition region of data at, af = t, f diff = 1e5 if transwidth != None: for i in t: d = abs(i - d50) if d < diff: mid = t.index(i) diff = d L = int(mid - transwidth) U = int(mid + transwidth) t, f = t[L:U], f[L:U] return at, af, t, f def fitVantHoff(self, E=None, d=None, xy=None, transwidth=80, invert=False, show=True, figname=None): """Derive fraction unfolded, get K and fit to Van't Hoff. see http://www.jbc.org/content/277/43/40717.full or http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2144003/ """ if E != None: if not d in E.datasets: print 'no such dataset, %s' % d print 'available datasets:', E.datasets return ek = E.getDataset(d) x, y = ek.getxySorted() elif xy != None: x, y = xy if 'Convert Celsius-Kelvin' in self.conversions.getcurselection(): x = [i + 273 for i in x] if invert == True: y = [max(y) - i for i in y[:]] f = plt.figure(figsize=(18, 6)) ax = f.add_subplot(131) p = ax.plot(x, y, 'o', alpha=0.6) ax.set_xlabel('T(K)') ax.set_ylabel('mdeg') ax.set_title('raw data') x1, y1, x, y = self.transformCD(x, y, transwidth, ax) cw = csv.writer(open('frac_unfolded_' + d + '.csv', 'w')) cw.writerow(['temp', 'frac']) for i in zip(x1, y1): cw.writerow(i) #derive lnK vs 1/T t = [] k = [] for T, fu in zip(x, y): if fu >= 1 or fu <= 0: continue K = fu / (1 - fu) klog = math.log(K) k.append(klog) t.append(1 / T) if len(t) < 2: return None, None, None ax = f.add_subplot(132) p = ax.plot(x1, y1, 'o', color='g', alpha=0.6) ax.set_xlabel('T(K)') ax.set_ylabel('fu') ax.set_title('fraction unfolded') ax = f.add_subplot(133) p = ax.plot(t, k, 'x', mew=2, color='black') ax.set_xlabel('1/T') #(r'$1/T ($K^-1)$') ax.set_ylabel('ln K') formatter = matplotlib.ticker.ScalarFormatter() formatter.set_scientific(True) formatter.set_powerlimits((0, 0)) ax.xaxis.set_major_formatter(formatter) for l in ax.get_xticklabels(): l.set_rotation(30) #fit this van't hoff plot A, X = Fitting.doFit(expdata=zip(t, k), model='Linear') fitk = X.getFitLine(t) p = ax.plot(t, fitk, 'r', lw=2) fd = X.getFitDict() #self.drawParams(ax,fd) #slope is deltaH/R/1000 in kJ deltaH = -fd['a'] * self.R / 1000 deltaS = fd['b'] * self.R / 1000 f.suptitle("Method 1 - deltaH: %2.2f deltaS: %2.2f" % (deltaH, deltaS), size=18) f.subplots_adjust(bottom=0.15, top=0.85) if show == True: self.showTkFigure(f) if figname == None: figname = d figname = figname.replace('.', '_') fname = figname + 'm1' + '.png' f.savefig(fname, dpi=300) print 'plot saved to %s' % os.path.abspath(fname) #plt.close() if E != None: fdata = Fitting.makeFitData(X.name, vrs=X.variables) E.insertDataset(xydata=[t, k], newname=d + '_vanthoff', replace=True, fit=fdata) #E.saveProject() return deltaH, deltaS, ax def fitElwellSchellman(self, E=None, d=None, xy=None, transwidth=50, invert=False, show=True, figname=None): """Fit entire raw data simultaneously to the three main thermodynamic parameters using Elwell/Schellman method""" if E != None: ek = E.getDataset(d) x, y, a, xerr, yerr = ek.getAll() elif xy != None: x, y = xy else: return if invert == True: y = [max(y) - i for i in y[:]] f = plt.figure(figsize=(10, 5)) ax = f.add_subplot(121) p = ax.plot(x, y, 'o', alpha=0.5) ax.set_xlabel('T') ax.set_xlabel('mdeg') ax.set_title('raw data') x1, y1, x, y = self.transformCD(x, y, transwidth, ax) t = [] dg = [] R = 8.3144e-3 for T, fu in zip(x, y): if fu >= 1 or fu <= 0: continue K = fu / (1 - fu) deltaGt = -R * T * math.log(K) dg.append(deltaGt) t.append(T) ax1 = f.add_subplot(122) p = ax1.plot(t, dg, 'x', mew=2, color='black') ax1.set_xlabel('T') ax1.set_ylabel('dG(T)') ax.set_title('stability curve') A, X = Fitting.doFit(expdata=zip(t, dg), model='schellman', grad=1e-9, conv=1e-9) fity = X.getFitLine(t) p = ax1.plot(t, fity, 'r', lw=2) fd = X.getFitDict() self.drawParams(ax1, fd) deltaH = fd['deltaH'] deltacp = fd['deltacp'] Tm = fd['Tm'] f.suptitle("Method 2 - deltaH: %2.2f deltaCp: %2.2e Tm: %2.2f" % (deltaH, deltacp, Tm), size=18) if show == True: self.showTkFigure(f) if figname == None: figname = d figname = figname.replace('.', '_') fname = figname + 'm1' + '.png' f.savefig(fname, dpi=300) print 'plot saved to %s' % os.path.abspath(fname) if E != None: fdata = Fitting.makeFitData(X.name, vrs=X.variables) E.insertDataset(xydata=[t, dg], newname=d + '_vanthoff2', replace=True, fit=fdata) #E.saveProject() return deltaH, Tm, deltacp def breslauerMethod(self, E=None, d=None, xy=None, invert=False, show=True, figname=None): """Finds slope of trans region and plugs this in to equation http://www.springerlink.com/content/r34n0201g30563u7/ """ if E != None: ek = E.getDataset(d) x, y, a, xerr, yerr = ek.getAll() elif xy != None: x, y = xy else: return f = plt.figure(figsize=(10, 6)) ax = f.add_subplot(111) ax.set_xlabel('T') p = ax.plot(x, y, 'o', alpha=0.5) d50 = self.guessMidpoint(x, y) A, X = Fitting.doFit(expdata=zip(x, y), model='Unfolding', conv=1e-7, noiter=60, guess=False, startvalues=[1, 1, 1, 1, 1, d50]) fity = X.getFitLine(x) p = ax.plot(x, fity, 'r', lw=2) fd = X.getFitDict() self.drawParams(ax, fd) Tm = fd['d50'] m = fd['m'] R = 8.3144e-3 deltaH = R * math.pow(Tm, 2) * m f.suptitle("Method 4 - deltaH: %2.2f Tm: %2.2f" % (deltaH, Tm), size=18) if show == True: self.showTkFigure(f) if figname != None: figname = figname.replace('.', '_') f.savefig(figname) plt.close() return deltaH, Tm def fitDifferentialCurve(self, E=None, d=None, xy=None, smooth=0, invert=False, show=True, figname=None): """Derive differential denaturation curve and fit to get deltaH We smooth the unfolding curve and then differentiate and finally fit to a 3 parameter equation. See http://www.ncbi.nlm.nih.gov/pubmed/10933511""" if E != None: ek = E.getDataset(d) x, y, a, xerr, yerr = ek.getAll() elif xy != None: x, y = xy else: return if invert == True: y = [max(y) - i for i in y[:]] leg = [] lines = [] f = plt.figure(figsize=(10, 5)) ax = f.add_subplot(121) p = ax.plot(x, y, 'x', color='black', mew=3, alpha=0.5) leg.append(p) lines.append('original') #smooth if smooth == 0: smooth = int(len(x) / 15.0) s = self.smoothListGaussian(y, smooth) p = ax.plot(x[:len(s) - 1], s[:-1], lw=3) leg.append(p) lines.append('smoothed') ax.set_title("original data") ax.set_xlabel('T') ax1 = f.add_subplot(122) #differentiate dx, ds = self.differentiate(x[:len(s)], s) #ds = [i/max(ds) for i in ds] ds = [i * 10 for i in ds] cw = csv.writer(open('diffcd.csv', 'w')) for row in zip(dx, ds): cw.writerow(row) p = ax1.plot(dx, ds, '-', lw=1.5, alpha=0.7, color='black') leg.append(p) lines.append('differential') ax1.set_title("differential denaturation") ax1.set_xlabel('T') ax1.set_ylabel('dsignal/dT') A, X = Fitting.doFit(expdata=zip(dx, ds), model='diffDenaturation', grad=1e-9, conv=1e-10) fity = X.getFitLine(dx) p = ax1.plot(dx, fity, 'r', lw=2) leg.append(p) lines.append('fit') t = X.getFitDict() self.drawParams(ax1, t) dHkcal = t['deltaH'] / 4.184 f.suptitle('Method 3 - deltaH: %2.2f kJ/mol (%2.2f kcal) Tm: %2.2f' % (t['deltaH'], dHkcal, t['Tm']), size=18) ax.legend(leg, lines, loc='best', prop=FontProperties(size="smaller")) #f.subplots_adjust(hspace=0.8) if show == True: self.showTkFigure(f) if figname != None: figname = figname.replace('.', '_') f.savefig(figname + 'm3', dpi=300) plt.close() if E != None: fdata = Fitting.makeFitData(X.name, vrs=X.variables) E.insertDataset(xydata=[dx, ds], newname=d + '_diff', replace=True, fit=fdata) #E.saveProject() return t['deltaH'], t['Tm'] def differentiate(self, x, y): dy = numpy.diff(y, 1) dx = x[:len(dy)] return dx, dy def smoothListGaussian(self, data, degree=5): """Gaussian data smoothing function""" #buffer data to avoid offset result data = list(data) data = [data[0]] * (degree - 1) + data + [data[-1]] * degree window = degree * 2 - 1 weight = numpy.array([1.0] * window) weightGauss = [] for i in range(window): i = i - degree + 1 frac = i / float(window) gauss = 1 / (numpy.exp((4 * (frac))**2)) weightGauss.append(gauss) weight = numpy.array(weightGauss) * weight smoothed = [0.0] * (len(data) - window) for i in range(len(smoothed)): smoothed[i] = sum( numpy.array(data[i:i + window]) * weight) / sum(weight) return smoothed def invert(self, data): inv = [i for i in data] return inv def simulateCD(self, noise=1.0): """Simulate some CD spec data""" x = list(numpy.arange(290, 380, 0.2)) y = [] X = Fitting.getFitter(model='Unfolding', vrs=[-16, 0.01, -11.6, 0.01, 2.7, 324]) fity = X.getFitLine(x) for i in fity: noise = numpy.random.normal(i, 1.0 / 2) y.append(i + noise) cw = csv.writer(open('cd.csv', 'w')) for row in zip(x, y): cw.writerow(row) return x, y def drawParams(self, ax, d): ymin, ymax = ax.get_ylim() xmin, xmax = ax.get_xlim() inc = (ymax - ymin) / 20 xinc = (xmax - xmin) / 20 y = ymax - inc for k in d: s = k + '=' + str(round(d[k], 3)) ax.text(xmin + xinc, y, s, fontsize=10) y -= inc return def pltConfig(self): #plt.rc('text', usetex=True) plt.rc('figure.subplot', hspace=0.3, wspace=0.3) #plt.rc('axes',titlesize=22) plt.rc('font', family='monospace') return def doAll(self, methods=['method 1']): """Process all datasets in ekinprj""" E = self.E vals = {} from Dialogs import PEATDialog pb = PEATDialog(self.mainwin, option='progressbar', message='Analysing Data..') pb.update_progress(0) total = len(E.datasets) count = 0 for d in E.datasets: if '_diff' in d or '_vanthoff' in d: continue vals[d] = {} name = d if 'method 1' in methods: vals[d]['dH1'], vals[d]['dS1'], ax = self.fitVantHoff( E, d, transwidth=int(self.tw.getvalue()), show=False, figname=name) if 'method 2' in methods: vals[d]['dH2'], vals[d]['dTm2'], vals[d][ 'dCp2'] = self.fitElwellSchellman(E, d, show=False, figname=name) if 'method 3' in methods: vals[d]['dH3'], vals[d]['dTm3'] = self.fitDifferentialCurve( E, d, show=False, figname=name) count += 1 pb.update_progress(float(count) / total * 100.0) pb.close() self.showTable(vals) return def showTable(self, data): """Show results in table""" from PEATDB.DictEdit import DictEditor D = DictEditor(self.mainwin) D.loadTable(data) return def benchmark(self, E=None, d=None, method=1): """Test methods with varying paramaters, smoothing etc""" if E == None and self.E != None: E = self.E d = self.dmenu.getcurselection() path = 'vh_benchmark' if not os.path.exists(path): os.mkdir(path) dHvals = [] if method == 1: xlabel = 'width (K)' title = 'method 1: deltaH variation with trans region width fit' vals = range(5, 140, 5) for w in vals: dH, dS, ax = self.fitVantHoff(E, d, transwidth=w, show=False, figname=os.path.join( path, '%s_%s.png' % (d, w))) if dH == None: dH = 0 dHvals.append(dH) #take best values from middle #dHvals= dHvals[5:16] elif method == 2: xlabel = 'width (K)' title = 'method 2: deltaH variation with width fit' vals = range(5, 140, 5) for w in vals: dH, dcp, dTm = self.fitElwellSchellman( E, d, transwidth=w, show=False, figname=os.path.join(path, '%s_%s.png' % (d, w))) dHvals.append(dH) elif method == 3: xlabel = 'smoothing degree' title = 'method 3: deltaH variation with degree of smoothing' vals = range(1, 30, 3) for s in vals: dH, dTm = self.fitDifferentialCurve(E, d, smooth=s, show=False, figname=os.path.join( path, '%s_%s.png' % (d, s))) dHvals.append(dH) mean = numpy.mean(dHvals) stdev = numpy.std(dHvals) f = plt.figure() ax = f.add_subplot(111) ax.plot(vals, dHvals, lw=2) ax.set_xlabel(xlabel) ax.set_ylabel('deltaH (kJ)') ax.set_title('mean: %2.2f stdev: %2.2f' % (mean, stdev)) f.suptitle(title) f.savefig('benchmark_%s.png' % method) cw = csv.writer(open('benchmark_%s.csv' % method, 'w')) for row in zip(vals, dHvals): cw.writerow(row) return def benchmarkLimitedData(self, E=None, d=None, method=1): """test any method with varying limited data""" if E == None and self.E != None: E = self.E d = self.dmenu.getcurselection() path = 'vh_benchmark' if not os.path.exists(path): os.mkdir(path) dHvals = [] vals = [] if method == 1: L = range(5, 140, 5) for w in vals: dH, dS, ax = self.fitVantHoff(E, d, transwidth=w, show=False, figname=os.path.join( path, '%s_%s.png' % (d, w))) return @classmethod def plotCorrelation(self, x=None, y=None, xlabel='method1', ylabel='method2'): if x == None: data = open('compared.csv', 'r') cr = csv.reader(data) x = [float(r[0]) for r in cr] data.seek(0) y = [float(r[1]) for r in cr] f = plt.figure() ax = f.add_subplot(111) line = ax.scatter(x, y, marker='o', alpha=0.8) cl = numpy.arange(0, max(x) + 50) ax.plot(cl, cl, 'g', alpha=0.5, lw=2) ax.set_xlabel(xlabel) ax.set_ylabel(ylabel) ax.set_xlim(150, 600) ax.set_ylim(150, 600) ax.set_title('Correlation') from scipy.stats import stats cc = str(round(pow(stats.pearsonr(x, y)[0], 2), 2)) ax.text(400, 180, r'$r^2= %s$' % cc, fontsize=16) self.showTkFigure(f) return def showTkFigure(self, fig): from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg, NavigationToolbar2TkAgg fr = Toplevel() canvas = FigureCanvasTkAgg(fig, master=fr) #self.canvas.show() canvas.get_tk_widget().pack(side=TOP, fill=X, expand=1) mtoolbar = NavigationToolbar2TkAgg(canvas, fr) mtoolbar.update() canvas._tkcanvas.pack(side=BOTTOM, fill=BOTH, expand=1) return