https://github.com/blab/antibody-response-pulse/
In [1]:
'''
author: Alvason Zhenhua Li
date: 04/09/2015
'''
%matplotlib inline
import numpy as np
import matplotlib.pyplot as plt
import os
AlvaFontSize = 23
AlvaFigSize = (9, 6)
numberingFig = 0
# plotting
dir_path = '/Users/al/Desktop/GitHub/antibody-response-pulse/bcell-array/figure'
file_name = 'Virus-Bcell-IgM-IgG'
figure_name = '-equation'
file_suffix = '.png'
save_figure = os.path.join(dir_path, file_name + figure_name + file_suffix)
numberingFig = numberingFig + 1
plt.figure(numberingFig, figsize=(12, 5))
plt.axis('off')
plt.title(r'$ Virus-Bcell-IgM-IgG \ equations \ (antibody-response \ for \ repeated-infection) $'
, fontsize = AlvaFontSize)
plt.text(0, 7.0/8, r'$ \frac{\partial V_n(t)}{\partial t} = \
+\mu_{v} V_{n}(t)(1 - \frac{V_n(t)}{V_{max}}) - \phi_{m} M_{n}(t) V_{n}(t) - \phi_{g} G_{n}(t) V_{n}(t) $'
, fontsize = 1.2*AlvaFontSize)
plt.text(0, 5.0/8, r'$ \frac{\partial B_n(t)}{\partial t} = \
+\mu_{b} + (\beta_{m} + \beta_{g}) V_{n}(t) B_{n}(t) - \mu_{b} B_{n}(t) $'
, fontsize = 1.2*AlvaFontSize)
plt.text(0, 3.0/8,r'$ \frac{\partial M_n(t)}{\partial t} = \
+\xi_{m} B_{n}(t) - \phi_{m} M_{n}(t) V_{n}(t) - \mu_{m} M_{n}(t) $'
, fontsize = 1.2*AlvaFontSize)
plt.text(0, 1.0/8,r'$ \frac{\partial G_n(t)}{\partial t} = \
+\xi_{g} B_{n}(t) - \phi_{g} G_{n}(t) V_{n}(t) - \mu_{g} G_{n}(t) $'
, fontsize = 1.2*AlvaFontSize)
plt.show()
# define the V-M-G partial differential equations
def dVdt_array(VBMGxt = [], *args):
# naming
V = VBMGxt[0]
B = VBMGxt[1]
M = VBMGxt[2]
G = VBMGxt[3]
x_totalPoint = VBMGxt.shape[1]
# there are n dSdt
dV_dt_array = np.zeros(x_totalPoint)
# each dSdt with the same equation form
for xn in range(x_totalPoint):
dV_dt_array[xn] = +inRateV*V[xn]*(1 - V[xn]/maxV) - killRateVm*M[xn]*V[xn] - killRateVg*G[xn]*V[xn]
return(dV_dt_array)
def dBdt_array(VBMGxt = [], *args):
# naming
V = VBMGxt[0]
B = VBMGxt[1]
M = VBMGxt[2]
G = VBMGxt[3]
x_totalPoint = VBMGxt.shape[1]
# there are n dSdt
dB_dt_array = np.zeros(x_totalPoint)
# each dSdt with the same equation form
for xn in range(x_totalPoint):
dB_dt_array[xn] = +inRateB + (actRateBm + actRateBg)*B[xn]*V[xn] - outRateB*B[xn]
return(dB_dt_array)
def dMdt_array(VBMGxt = [], *args):
# naming
V = VBMGxt[0]
B = VBMGxt[1]
M = VBMGxt[2]
G = VBMGxt[3]
x_totalPoint = VBMGxt.shape[1]
# there are n dSdt
dM_dt_array = np.zeros(x_totalPoint)
# each dSdt with the same equation form
for xn in range(x_totalPoint):
dM_dt_array[xn] = +inRateM*B[xn] - consumeRateM*M[xn]*V[xn] - outRateM*M[xn]
return(dM_dt_array)
def dGdt_array(VBMGxt = [], *args):
# naming
V = VBMGxt[0]
B = VBMGxt[1]
M = VBMGxt[2]
G = VBMGxt[3]
x_totalPoint = VBMGxt.shape[1]
# there are n dSdt
dG_dt_array = np.zeros(x_totalPoint)
# each dSdt with the same equation form
for xn in range(x_totalPoint):
dG_dt_array[xn] = +inRateG*B[xn] - consumeRateG*G[xn]*V[xn] - outRateG*G[xn]
return(dG_dt_array)
# define RK4 for an array (3, n) of coupled differential equations
def AlvaRungeKutta4ArrayXT(pde_array, startingOut_Value, minX_In, maxX_In, totalGPoint_X, minT_In, maxT_In, totalGPoint_T):
# primary size of pde equations
outWay = pde_array.shape[0]
# initialize the whole memory-space for output and input
inWay = 1; # one layer is enough for storing "x" and "t" (only two list of variable)
# define the first part of array as output memory-space
gridOutIn_array = np.zeros([outWay + inWay, totalGPoint_X, totalGPoint_T])
# loading starting output values
for i in range(outWay):
gridOutIn_array[i, :, :] = startingOut_Value[i, :, :]
# griding input X value
gridingInput_X = np.linspace(minX_In, maxX_In, num = totalGPoint_X, retstep = True)
# loading input values to (define the final array as input memory-space)
gridOutIn_array[-inWay, :, 0] = gridingInput_X[0]
# step-size (increment of input X)
dx = gridingInput_X[1]
# griding input T value
gridingInput_T = np.linspace(minT_In, maxT_In, num = totalGPoint_T, retstep = True)
# loading input values to (define the final array as input memory-space)
gridOutIn_array[-inWay, 0, :] = gridingInput_T[0]
# step-size (increment of input T)
dt = gridingInput_T[1]
# starting
# initialize the memory-space for local try-step
dydt1_array = np.zeros([outWay, totalGPoint_X])
dydt2_array = np.zeros([outWay, totalGPoint_X])
dydt3_array = np.zeros([outWay, totalGPoint_X])
dydt4_array = np.zeros([outWay, totalGPoint_X])
# initialize the memory-space for keeping current value
currentOut_Value = np.zeros([outWay, totalGPoint_X])
for tn in range(totalGPoint_T - 1):
# setting virus1 = 0 if virus1 < 1
if gridOutIn_array[0, 0, tn] < 1.0:
gridOutIn_array[0, 0, tn] = 0.0
# keep initial value at the moment of tn
currentOut_Value[:, :] = np.copy(gridOutIn_array[:-inWay, :, tn])
currentIn_T_Value = np.copy(gridOutIn_array[-inWay, 0, tn])
# first try-step
for i in range(outWay):
for xn in range(totalGPoint_X):
dydt1_array[i, xn] = pde_array[i](gridOutIn_array[:, :, tn])[xn] # computing ratio
gridOutIn_array[:-inWay, :, tn] = currentOut_Value[:, :] + dydt1_array[:, :]*dt/2 # update output
gridOutIn_array[-inWay, 0, tn] = currentIn_T_Value + dt/2 # update input
# second half try-step
for i in range(outWay):
for xn in range(totalGPoint_X):
dydt2_array[i, xn] = pde_array[i](gridOutIn_array[:, :, tn])[xn] # computing ratio
gridOutIn_array[:-inWay, :, tn] = currentOut_Value[:, :] + dydt2_array[:, :]*dt/2 # update output
gridOutIn_array[-inWay, 0, tn] = currentIn_T_Value + dt/2 # update input
# third half try-step
for i in range(outWay):
for xn in range(totalGPoint_X):
dydt3_array[i, xn] = pde_array[i](gridOutIn_array[:, :, tn])[xn] # computing ratio
gridOutIn_array[:-inWay, :, tn] = currentOut_Value[:, :] + dydt3_array[:, :]*dt # update output
gridOutIn_array[-inWay, 0, tn] = currentIn_T_Value + dt # update input
# fourth try-step
for i in range(outWay):
for xn in range(totalGPoint_X):
dydt4_array[i, xn] = pde_array[i](gridOutIn_array[:, :, tn])[xn] # computing ratio
# solid step (update the next output) by accumulate all the try-steps with proper adjustment
gridOutIn_array[:-inWay, :, tn + 1] = currentOut_Value[:, :] + dt*(dydt1_array[:, :]/6
+ dydt2_array[:, :]/3
+ dydt3_array[:, :]/3
+ dydt4_array[:, :]/6)
# restore to initial value
gridOutIn_array[:-inWay, :, tn] = np.copy(currentOut_Value[:, :])
gridOutIn_array[-inWay, 0, tn] = np.copy(currentIn_T_Value)
# end of loop
return (gridOutIn_array[:-inWay, :])
In [2]:
# Experimental lab data from OAS paper
gT_lab = np.array([0, 5, 10, 20, 25, 30, 60, 80])
gIgG_lab = np.array([0, 0, 4, 8, 8.5, 7.5, 5, 4])*10**2
gIgM_lab = np.array([0, 0.2, 2.5, 0.2, 0.1, 0, 0, 0])*10**2
# setting parameter
timeUnit = 'day'
if timeUnit == 'hour':
hour = float(1); day = float(24);
elif timeUnit == 'day':
day = float(1); hour = float(1)/24;
maxV = float(1000) # max virus/milli-liter
inRateV = 6.5*maxV/10**4 # in-rate of virus
killRateVm = 1*maxV/10**5 # kill-rate of virus by antibody-IgM
killRateVg = killRateVm/1 # kill-rate of virus by antibody-IgG
inRateB = 2*maxV/10**4 # in-rate of B-cell
outRateB = inRateB # out-rate of B-cell
actRateBm = killRateVm # activation rate of naive B-cell
actRateBg = killRateVg # activation rate of memory B-cell
inRateM = maxV*1/10**2 # in-rate of antibody-IgM from naive B-cell
outRateM = inRateM # out-rate of antibody-IgM from naive B-cell
consumeRateM = killRateVm # consume-rate of antibody-IgM by cleaning virus
inRateG = inRateM/20 # in-rate of antibody-IgG from memory B-cell
outRateG = outRateM/600 # out-rate of antibody-IgG from memory B-cell
consumeRateG = killRateVg # consume-rate of antibody-IgG by cleaning virus
# time boundary and griding condition
minT = float(0)
maxT = float(80*day)
totalGPoint_T = int(10**4 + 1)
gridT = np.linspace(minT, maxT, totalGPoint_T)
spacingT = np.linspace(minT, maxT, num = totalGPoint_T, retstep = True)
gridT = spacingT[0]
dt = spacingT[1]
# space boundary and griding condition
minX = float(0)
maxX = float(1)
totalGPoint_X = int(1 + 1)
gridX = np.linspace(minX, maxX, totalGPoint_X)
gridingX = np.linspace(minX, maxX, num = totalGPoint_X, retstep = True)
gridX = gridingX[0]
dx = gridingX[1]
gridV_array = np.zeros([totalGPoint_X, totalGPoint_T])
gridB_array = np.zeros([totalGPoint_X, totalGPoint_T])
gridM_array = np.zeros([totalGPoint_X, totalGPoint_T])
gridG_array = np.zeros([totalGPoint_X, totalGPoint_T])
# initial output condition
gridV_array[0, 0] = float(1)
gridB_array[0, 0] = float(0)
gridM_array[0, 0] = float(0)
gridG_array[0, 0] = float(0)
# Runge Kutta numerical solution
pde_array = np.array([dVdt_array, dBdt_array, dMdt_array, dGdt_array])
startingOut_Value = np.array([gridV_array, gridB_array, gridM_array, gridG_array])
gridOut_array = AlvaRungeKutta4ArrayXT(pde_array, startingOut_Value, minX, maxX, totalGPoint_X, minT, maxT, totalGPoint_T)
# plotting
gridV = gridOut_array[0]
gridB = gridOut_array[1]
gridM = gridOut_array[2]
gridG = gridOut_array[3]
figure_name = '-first-infection'
figure_suffix = '.png'
save_figure = os.path.join(dir_path, file_name + figure_name + file_suffix)
numberingFig = numberingFig + 1
for i in range(1):
plt.figure(numberingFig, figsize = AlvaFigSize)
plt.plot(gridT, gridV[i], color = 'red', label = r'$ V_{%i}(t) $'%(i), linewidth = 3.0, alpha = 0.5)
plt.plot(gridT, gridM[i], color = 'blue', label = r'$ IgM_{%i}(t) $'%(i), linewidth = 3.0, alpha = 0.5)
plt.plot(gridT, gridG[i], color = 'green', label = r'$ IgG_{%i}(t) $'%(i), linewidth = 3.0, alpha = 0.5)
plt.plot(gridT, gridM[i] + gridG[i], color = 'gray', linewidth = 5.0, alpha = 0.5, linestyle = 'dashed'
, label = r'$ IgM_{%i}(t) + IgG_{%i}(t) $'%(i, i))
plt.plot(gT_lab, gIgG_lab, marker = 'o', markersize = 20, alpha = 0.3, color = 'green', label = r'$ IgG-(lab-X31) $')
plt.plot(gT_lab, gIgM_lab, marker = 's', markersize = 20, alpha = 0.3, color = 'blue', label = r'$ IgM-(lab-X31) $')
plt.grid(True, which = 'both')
plt.title(r'$ Antibody \ for \ First-infection $', fontsize = AlvaFontSize)
plt.xlabel(r'$time \ (%s)$'%(timeUnit), fontsize = AlvaFontSize)
plt.ylabel(r'$ Serum \ antibody \ (pg/mL) $', fontsize = AlvaFontSize)
plt.xticks(fontsize = AlvaFontSize*0.6)
plt.yticks(fontsize = AlvaFontSize*0.6)
plt.text(maxT*16.0/10, maxV*8.0/10, r'$ V_{max} = %f $'%(maxV), fontsize = AlvaFontSize)
plt.text(maxT*16.0/10, maxV*7.0/10, r'$ \mu_{v} = %f $'%(inRateV), fontsize = AlvaFontSize)
plt.text(maxT*16.0/10, maxV*6.0/10, r'$ \phi_{m} = %f $'%(killRateVm), fontsize = AlvaFontSize)
plt.text(maxT*16.0/10, maxV*5.0/10, r'$ \phi_{g} = %f $'%(killRateVg), fontsize = AlvaFontSize)
plt.text(maxT*16.0/10, maxV*4.0/10, r'$ \mu_{b} = %f $'%(inRateB), fontsize = AlvaFontSize)
plt.text(maxT*16.0/10, maxV*3.0/10, r'$ \xi_{m} = %f $'%(inRateM), fontsize = AlvaFontSize)
plt.text(maxT*16.0/10, maxV*2.0/10, r'$ \xi_{g} = %f $'%(inRateG), fontsize = AlvaFontSize)
plt.text(maxT*16.0/10, maxV*1.0/10, r'$ \mu_{g} = %f $'%(outRateG), fontsize = AlvaFontSize)
plt.ylim(-100, 1000)
plt.legend(loc = (1, 0), fontsize = AlvaFontSize)
plt.savefig(save_figure, dpi = 100, bbox_inches='tight')
plt.show()
In [2]: