ECCpy usage snapshot

Here a dataset is examined that consists of four samples and a control. An LD50 experiment is repeated over three days. (code for generating this example is shown below)


The data for each day is prepared in an excel file, to be used as ECCpy input. The excel template contains three tabs. 1) dose, 2) response, and 3) a "template_identifier" tab that helps ECCpy identify the excel format.

An excel settings file controls all ECCpy inputs, including the list of files to examine. Update the settings file to include data files.

Run curve_fit in python or the ipython/jupyter notebook, to fit a sigmoidal curve to the dose-response data, and calculate EC50 values.

import eccpy
settings = r"D:\data\ECCpy_settings.xlsx"
eccpy.run_curvefit(settings)

The fitted curves are shown in the .png images. The list of calculated EC50 values are saved in excel and csv.

A summary figure for each day/experiment is created. Curve-fitting is accompanied by a judgefit program, that examines the quality of the curve fitting and EC50 value. Samples with a red letter code in the barchart x-axis (e.g. AB, sample 2) have been automatically judged by ECCpy as low quality..

In the list of quality filters on the right-hand-side, the red "r_squared" and "std_highdose_datapoints" show that for these filters, the fitted curve did not meet the user-defined quality threshold (as defined in the settings excel file). If the data fails to meet any of the thresholds, the EC50 is marked as red, and is excluded from the analysis program that compares data from multiple experiments. The curve is shown with both the original x- and y-axes (top left) and with normalised axes between 0 and 1 (bottom left).

In some cases, the the curve itself is annotated in red to show why the data quality is judged to be low. For example here the slope at the last datapoint is too high. Repeating the experiment with higher dose concentrations would be necessary.

The results from all three days can be compared with the run_analysis program. First mark "run_analysis" in the excel settings file as "TRUE":

run_analysis in python or the ipython/jupyter notebook.

import eccpy
settings = r"D:\data\ECCpy_settings.xlsx"
eccpy.run_gatherer(settings)

The analysis program combines all high-quality EC50 values from the selected experiments, and presents the data as a barchart and scatter plot.

The compare_raw program goes through all the EC50 values calculated for the selected experiments, and compares the raw data curves for the selected samples.

# define which samples you want to compare, across all experiments
samples = ["control", "sample1"]
eccpy.compare_rawdata(settings_excel_file, samples)

Code to generate an example dataset for analysis.

import eccpy
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
% matplotlib inline
t20 = eccpy.tools.setup_t20_colour_list()
dose_values = np.linspace(0,1,24)
days = ["day1", "day2", "day3"]
samples = ["sample1", "sample2","control","sample3","sample4"]
df_dose = pd.DataFrame()
df_response = pd.DataFrame()
for i, day in enumerate(days):
    fig, ax = plt.subplots()
    ax.set_title(day)
    for n, sample in enumerate(samples):
        # add dose values
        df_dose[sample] = dose_values
        # create the hill constants for curves with EC50 offset and some randomisation
        random_offset = n*0.18 + np.random.randn()/20
        EC50 = 0.2 + random_offset
        if day == "day3":
            EC50 = EC50 - 0.1 
        hill_constants = [100, 0, EC50, 7.1]
        # generate perfect response curve
        response_curve = eccpy.tools.hill_eq(hill_constants, dose_values)
        # scatter datapoints to resemble real data. Add to dataframe and plot.
        random = np.random.random(len(dose_values)) * 10
        response_values = response_curve + random
        if day == "day1" and sample == "sample2":
            response_values[21] = (response_values.max() - response_values.min())*0.95
            response_values[22] = (response_values.max() - response_values.min())*0.95
        if day == "day2" and sample == "sample1":
            response_values[::2] = response_values[::2]*1.2 + response_values.max()/5
        if day == "day2" and sample == "sample4":
            response_values[-1] = 0
        #if day == "day2" and sample == "sample2":
        #    response_values = response_values + random
        df_response[sample] = response_values
        ax.scatter(dose_values, df_response[sample], color=t20[n],label=sample)
    ax.legend(loc="best")
    ax.set_xlim(-0.1,1.5)
    writer = pd.ExcelWriter("{}_{}.xlsx".format(day, i))
    df_dose.to_excel(writer, sheet_name="dose", index=False)
    df_response.to_excel(writer, sheet_name="response", index=False)
    template = "eccpy|xlsx|generic|vertical"
    template_identifier = pd.DataFrame(template,[0], [0])
    template_identifier.to_excel(writer, sheet_name="template_identifier", index=False, header=False)
    writer.save()
    writer.close()