Example analysis of cleavage rate dispersion model#
Below is an example of how to load and visualize a dispersion model creates using learn_dm.
Load dispersion model file#
Here we remotely load the the dispersion model created for human regulatory T cells
[1]:
from footprint_tools.modeling import dispersion
dm_filepath = 'https://resources.altius.org/~jvierstra/projects/footprinting.2020/per.dataset/hTreg-DS14702/dm.json'
dm = dispersion.load_dispersion_model(dm_filepath)
Plot overall model parameters#
The negative binomial distrubution has two parameters: mean and variance. The dispersion model is constructed by performing a MLE fit for each predicted cleavage rate and then smoothing the parameters using a piece-wise regression. The code below plots the MLE fit paramters (blue lines) and the piece-wise regression (dashed orange).
[2]:
from footprint_tools.plotting import plot_model_fit
fig, axs = plot_model_fit(dm)
fig.set_size_inches(8, 2.5)
Plot fit for a predicted cleavage rate#
The code below plots the probability density for a single predicted cleavage rate. The negative binomial distribution shown (orange) is used in the statistical tests for cleavage deviation (lower-tail p-value) and for footprint likelihood functions (probability density function). A Poisson distrubution is additionally plotted to demonstrate overdispersion.
[3]:
from footprint_tools.plotting import plot_model_histogram
fig, ax = plot_model_histogram(dm, n=100, xlim=(0, 250))
fig.set_size_inches(4, 2.5)
