Pyvivc

Version 2.0

Stephen Checkley, October 2021.

Description

A Python 3 port of the Rivivc R package for IVIVC linear level A by Aleksander Mendyk and Sebastian Polak. The package contains a numerical deconvolution method working for inequal and incompatible timepoints between impulse and response curves. A numerical convolution method is also included.

This version faithfully reproduces the algorithm of Rivivc 0.9 and has no pandas dependency — curves are passed as plain numpy arrays with two columns: column 0 = time, column 1 = concentration.

Installation

Clone the repository and install with pip:

pip install .

or install with pip from the PyPi repository:

pip install pyvivc

Pyvivc example

from pyvivc import *
import numpy as np
import matplotlib.pyplot as plt
import matplotlib
matplotlib.use('TkAgg')


def load_curve(path):
    # skip the "time,C" header; return an (n, 2) float array
    return np.genfromtxt(path, delimiter=',', skip_header=1)


impulse = load_curve('data/impulse.csv')
response = load_curve('data/resp.csv')
inp = load_curve('data/input.csv')

out = pyivivc(inp, impulse, response,
              explicit_interpolation=10, implicit_interpolation=5)

regression = out[0]   # scipy linregress result
numeric = out[1]      # numpy array: col 0 = time, col 1 = par

x = inp[:, 1]         # input dissolution (fraction)
y = numeric[:, 1]     # deconvolved input

rsquare_text = 'R squared = ' + str(round(regression.rvalue, 2))

plt.subplot(1, 2, 1)
plt.plot(x, y, 'o', label='data')
plt.plot(y, regression.intercept + regression.slope * y, 'r')
plt.annotate(rsquare_text, (0, 0.8), horizontalalignment='left',
             verticalalignment='top', fontsize=8)
plt.xlabel('input data (#)')
plt.ylabel('deconvolved input (#)')
plt.legend()

plt.subplot(1, 2, 2)
plt.plot(numeric[:, 0], y, 'r', label='deconvolution')
plt.plot(inp[:, 0], inp[:, 1], 'o', label='data')
plt.xlabel('Time')
plt.ylabel('discovered input (%)')
plt.legend()

plt.show()

API

All curves are numpy arrays of shape (n, 2): column 0 = time, column 1 = concentration.

• NumConv(impulse, input, conv_timescale=None, explicit_interpolation=1000) returns a dict with keys par and par_explicit (each an (n, 2) numpy array).
• NumDeconv(impulse, response, dose_iv=None, dose_po=None, deconv_timescale=None, explicit_interpolation=20, implicit_interpolation=10, maxit_optim=200) returns a dict with keys par, par_explicit, and par_implicit.
• pyivivc(known_dat, impulse, second_profile, ...) returns [scipy_linregress_result, numeric_array].