htmek

General utilities for HT-MEK data processing with magnify.

Installation

Install pip, then pip install magnify and PyQt5

conda install pip
pip install magnify pyqt5 scikit-image

Next, clone and install htmek from this repository

git clone https://github.com/FordyceLab/htmek.git
cd htmek
pip install -e .

Processing Pipelines

Buttons

Subtract egfp background

htmek.utils.subtract_bg(
    path/to/final_egfp_image,
    path/to/background_egfp_image,
    overlap = 82, # use 82 for 0.08 overlap, 102 for 0.1 overlap
    output = 'egfp_bg_sub.tif'
)

Button finding and analysis

# Make a chip object
egfp_chip = htmek.buttons_pipe(
    'egfp_bg_sub.tif',
    path/to/pinlist,
    overlap = 0, # assuming using pre-stitched bg image from above
)

# Quickly view the chip interactively
htmek.viz.view(egfp_chip, limits=(1000, 10000))

# Look at each chamber individually with mask
htmek.viz.chamber_map(egfp_chip, limits=(1000, 16000))

# Look at all chambers for one variant/pinlist tag
htmek.viz.chamber_map(egfp_chip, tag='N41A', limits=(1000, 16000))

# Make a heatmap of all summed button intensitites
htmek.viz.chip_hm(egfp_chip, agg_func='sum')

Get dataframe of summed egfp intensity

# Convert to dataframe
egfp_df = htmek.utils.to_df(egfp_chip, agg_func='sum')

# Rename generic 'roi' column
egfp_df = egfp_df.rename({'roi': 'summed_egfp'}, axis=1)

# Export to csv
egfp_df.to_csv('egfp_data.csv', index=False)

Checking assay progress curves

It can be beneficial to check progress curves as they are collected to ensure that the data look as expected/high quality, prior to collected stanard curves to make the results more quantitative. These pipelines are similar to those used for analyzing standards and properly analyzing kinetics data below.

Fitting process

Fitting data using htmek requires a fit_parameters variable, which is just a special namedtuple contaning information about how to fit the data and what exactly is being fit.

The main goal of this is to create flexible but well-defined processes by which we can fit data and retain information about how the function was fit.

A simple example for fitting a line is given below:

# First define the function you are fitting
def linear(x, m, b):
    """Fits a line given the slope and y-intercept."""
    return m*x+b

# Pack this into `fit_parameters` with additonal info
linear_fit_params = fit_parameters(
    'linear',
    linear,
    param_names = ['m', 'b'],
    p0 = [1, 0], 
    param_bounds = ([0]*2, [np.inf]*2),
    xlimits = 50,
    ylimits = None,
)

# Fit the data
fit_dict, fit_map = htmek.process.fit(
    df = df,
    fit_params = [linear_fit_params],
    x_label = 'time (s)',
    y_label = 'median intensity',
)

Description of each parameter:

fit_params = fit_parameters(
    name: name of the function
    func: the function itself
    param_names: names for each fit parameter
    p0: initial guesses for each parameter (see scipy.optimize.curve_fit)
    param_bounds: ounds for each parameter (see scipy.optimize.curve_fit)
    xlimits: limits on the x-data used for the fit. Can be None, int (upper limit), or tuple (lower, upper), and can contain np.infs.
    ylimits: same as xlimits, but for y-data.
)