def fitToModel(tPoints, ratios, times, turnover, tor_v, rFile, numToFit, BGD, numToSample, alpha):
# Assign model parameters
time_sec = [x * 60 for x in tPoints]
time = [x * 1 for x in tPoints]
#time = time_sec # Turn-over ratio sample time-points (in seconds)
proteinRel = ratios[:]
proteinTime = times[:]
par_mod = list()
# create list of parameters to send into function
par_mod.append(time_sec); # Turn-over ratio sample time-points (in seconds)
par_mod.append(proteinTime) # Protein times from input file
par_mod.append(proteinRel) # Relative protein level
par_mod.append(0) # initial concentration
par_mod.append(0.0) # background
R_ccc_all = list()
R_ccc_all_best = list()
lin_turn_best = list()
# for each site in the turnover matrix
num = 0
ssdnum = 0
for s in range(0, len(turnover)):
#plt.figure(num)
name = str(turnover[s][0]) #+ '\t'+ str(turnover[s][1]) + '\t' + str(turnover[s][2]) + '\t'+ str(turnover[s][3])
SSDs = list()
acceptedSSDs = list()
print('Fitting model to data for site: ' + name)
#k = 0
lin_turn = list() # linear form of turnover data for this site
for c in range(1,len(turnover[s])):
lin_turn.append(pow(2, turnover[s][c]))
num+=1
SSD = 1E15; # start with a really big sum of squares difference
bestP = tor_v[0]
val_init = bestP
#tor_v = [1.010261e-04]
for p in tor_v: # for each initial parameter choice, p
# sample a range of parameters
newP = p
# run model with new parameter & calculated SSD
R_ccc = model_v0(newP, par_mod, lin_turn, BGD)
#plt.plot(time,R_ccc, 'g-*')
newSSD = 0
# check to see if the model was solved:
if (len(R_ccc) >= (numToFit)):
# fit to time points in
for v in range(0,numToFit):
newSSD += pow((lin_turn[v] - R_ccc[v]), 2)
SSDs.append(newSSD)
else: # model did not finish
newSSD = -1
#print('SSD: ' + str(newSSD))
#if (newSSD < SSD):
# bestP = newP # set this as the new best parameter
# SSD = newSSD
# val_init = p
# acceptedSSDs.append(newSSD)
# only sample around this parameter if this SSD is better than best
if (newSSD > 0): #newSSD <= SSD):
#print('old ssd:' + str(SSD) + ' new ssd: ' + str(newSSD))
for i in range(0,numToSample):
# sample new parameter
newP = random.uniform(newP-newP*alpha, newP+newP*alpha)
while (newP < 0):
newP = random.uniform(newP-newP*alpha, newP+newP*alpha)
# run the model
R_ccc = model_v0(newP, par_mod, lin_turn, BGD)
#plt.plot(time,R_ccc, 'g-*')
newSSD = 0
if (len(R_ccc) >= (numToFit)):
for v in range(0,numToFit):
newSSD = newSSD + pow(R_ccc[v]-lin_turn[v], 2)
SSDs.append(newSSD)
else:
newSSD = -1
if ((newSSD < SSD) and (newSSD != -1)):
#print('selecting new SSD')
#for v in range(0,numToFit):
# print(str(R_ccc[v]))
# print(str(lin_turn[v]))
bestP = newP # set this as the new best parameter
SSD = newSSD
val_init = p
acceptedSSDs.append(newSSD)
R_ccc_best = model_v0(bestP, par_mod, lin_turn, BGD)
R_ccc_all_best.append(R_ccc_best)
lin_turn_best.append(lin_turn)
R_ccc_all.append([name, val_init, bestP, SSD])
ssdnum = ssdnum + 1
num = num + 1
# write results to file
print ('Writing model fit results to file....')
file = open(rFile, 'w')
# print header line
header = 'Site_Chr' + '\t' + 'Site_Pos' + '\t' + 'Mapped_Chr' + '\t' + 'Mapped_Pos' + '\t' + 'initial turnover' + '\t' + 'predicted turnover' + '\t' + 'best SSD' + '\t'
for t in range(0, len(tPoints) - 1):
header = header + str(tPoints[t]) + '\t'
header = header + str(tPoints[len(tPoints)-1]) + '\n'
file.write(header)
i = 0
for result in R_ccc_all: # for each row
file.write(result[0] + '\t' + str(result[1]) + '\t' + str(result[2]) + '\t' + str(result[3]) +'\t')
for j in range(0,len(R_ccc_all_best[i])-1):
# write log 2 transformed
file.write(str(log(R_ccc_all_best[i][j],2)) + '\t')
file.write(str(log(R_ccc_all_best[i][len(R_ccc_all_best[i])-1],2)) +'\n')
i += 1
file.close()
return [R_ccc_all_best, lin_turn_best]
def plot_multiple_fits(turnover, resultsFileArray, numToFitArray, ratios, times, tPoints, figName):
plt.ioff() # turn off interactive plotting
# Assign model parameters
time_sec = [x * 60 for x in tPoints]
time = [x * 1 for x in tPoints] # Turn-over ratio sample time-points (in seconds)
proteinRel = ratios[:]
proteinTime = times[:]
# initialize absolute protein
proteinAbs = list()
for value in ratios:
proteinAbs.append(value/(1-value))
par_mod = list()
# create list of parameters to send into function
par_mod.append(time_sec); # Turn-over ratio sample time-points (in seconds)
par_mod.append(proteinTime) # Protein times from input file
par_mod.append(proteinRel) # Relative protein level
par_mod.append(0) # initial concentration
par_mod.append(0.0) # background
# number of background points to use
BGD = 2
# create array of colors for plots
colors = ['k-*', 'r', 'y', 'g', 'b']
SSDColors = ['k', 'ro', 'yo', 'go', 'bo']
# create array to hold file streams
fileStream = list()
# site number = line number in file stream
lineNum = 1 # first line in file is a header line
print ('Processing the following files:')
for file in resultsFileArray:
print(file)
# for each peak in turnover2D
for s in range(0, len(turnover)):
lineNum = lineNum + 1 # line number in the results file
# set up plot color index
colIndex = 0
# initialize list to hold turnover data for this site
lin_turn = list()
# pull site name
name = turnover[s][0]
# linearize turnover data
for c in range(1,len(turnover[s])):
lin_turn.append(pow(2, turnover[s][c]))
# create figure to hold turnover data and model fits plot
fig1 = plt.figure('Model Fits for ' + name, figsize=(10,10))
# plot linearized version of turnover data
plt.plot(time, lin_turn, colors[colIndex], linewidth=3, label = 'exp', markersize=15)
# create figure to hold ssd v. turnover rate plot
fig2 = plt.figure('Residence Time vs. SSD for ' + name, figsize=(10,10))
# for each open file
for file in resultsFileArray:
# increment color index
colIndex = colIndex + 1
# read current line
site = linecache.getline(file, lineNum)
# split current line and save turnover rate
turnoverRate = float(site.split('\t')[1])
SSD = float(site.split('\t')[2])
# plot turnover rate as function of SSD
plt.figure('Residence Time vs. SSD for ' + name)
plt.plot(1/turnoverRate, SSD, SSDColors[colIndex], markersize=15, label='# points = ' + str(numToFitArray[colIndex-1]))
# run model using turnover rate
R_ccc = model_v0(turnoverRate, par_mod, lin_turn, BGD)
# plot mode fit for this turnover rate
plt.figure('Model Fits for ' + name)
plt.plot(time, R_ccc, colors[colIndex], linewidth=1.5, label='# pts = ' + str(numToFitArray[colIndex-1]) + ' RT = ' + "%.0f" % (1/turnoverRate))
#str(1/turnoverRate) )
# add axis labels to model fit plot
plt.xlabel('Time (min)')
plt.ylabel('Turnover ratio (linear)')
plt.title(name)
plt.ylim(min(lin_turn), max(lin_turn)*1.5)
plt.legend(loc=2, fontsize=10)
# save model fit plot
plt.savefig(figName + '_Fits_' + name)
plt.close(fig1)
# add axis labels to ssd vs. turnover rate plot
plt.figure('Residence Time vs. SSD for ' + name) # select this figure
plt.xlabel('Predicted Residence Time (RT)')
plt.ylabel('Sum of squares difference (SSD)')
plt.title(name + ' SSD')
plt.legend(loc=2, fontsize=10)
# save ssd vs turnover rate plot
plt.savefig(figName + '_SSDvTO_' + name)
plt.close(fig2)
# for each file in resultsFileArray
for stream in fileStream:
# close each file
stream.close()
