Implemented classes¶

Autoencoder¶

Matrix¶

class src.matrix.Hic(cooler, *args, **kwargs)[source]¶

class Hic¶
This class inherits the Matrix class and set the matrix numpy array for a Hi-C data.

cooler¶

Storage of the Hi-C matrix

Type:	cooler

calculate_cum_length()[source]¶

Calculates and returns the cumulated length from chromosome 1 to N.

Returns:	Informations on chromosomes, their length and cumulated length
Return type:	Pandas DataFrame

set_matrix()[source]¶: Set the Hi-C numpy array of the chromosome chrom_num. The matrix is transformed into an upper triangular matrix and the values are converted in float32 and rescaled by log10 and normalized.

class src.matrix.HistoneMark(bed_file, *args, **kwargs)[source]¶

class HistoneModification¶
This class inherits the Matrix class and set the matrix numpy array for a histone mark.

mark_df¶

Histone modification sparse matrix

Type:	Pandas Dataframe

set_matrix()[source]¶: Set the histone modification numpy array of the chromosome chrom_num. The values of the matrix are converted in float32 and rescaled by log10 and normalized.

class src.matrix.Matrix(resolution, chrom_num, side)[source]¶

class Matrix¶
This class stores a matrix and different related numpy array, plots and writes this matrix.

resolution¶

Resolution (or bin size) of the matrix

Type:	int

chrom_num¶

Chromosome chosen for processing

Type:	int

side¶

Square side of a numpy array sub-matrix

Type:	int

matrix¶

Matrix stored in a numpy array

Type:	numpy array

sub_matrices¶

The matrix is divided into S sub-matrices of size side*side and stored in a numpy array of shape (X, side, side, 1)

Type:	numpy array

white_sub_matrices_ind¶

Position of the blank sub-matrices

Type:	list

total_sub_matrices¶

Total number of sub-matrices

Type:	int

latent_spaces¶

Latent spaces (encoded sub-matrices) stored in a numpy array

Type:	numpy array

predicted_sub_matrices¶

Predicted sub_matrices (decoded latent spaces) stored in a numpy array

Type:	numpy array

plot_distribution_matrix(matrix_type, path)[source]¶

Plot the distribution of the matrix.

Parameters:	matrix_type (str) – Matrix’s name path (str) – Path of the output plot

plot_matrix(matrix_type, color_map, path)[source]¶

The matrix is plotted in a file.

Parameters:	matrix_type (str) – Matrix’s name color_map (matplotlib.colors.ListedColormap) – Color map path (str) – Path of the output plot

plot_sub_matrices(matrix_type, index_list, color_map, path)[source]¶

40 random sub-matrices are plotted in a file.

Parameters:	matrix_type (str) – Matrix’s name index_list (list) – List of the 40 sub-matrix indexes to plot color_map (matplotlib.colors.ListedColormap) – Color map path (str) – Path of the output plot

set_predicted_latent_spaces(latent_spaces)[source]¶

Set the latent spaces predicted by the encoder.

Parameters:	latent_spaces (numpy array) – The predicted latent_spaces

set_predicted_sub_matrices(predicted_sub_matrices)[source]¶

Set the sub-matrices predicted by the whole autoencoder.

Parameters:	predicted_sub_matrices (numpy array) – The predicted sub-matrices

set_sub_matrices()[source]¶: Divide the matrix into S sub-matrices of size side*side. The empty sub-matrices (sum(values)==0) are removed from the data set. The S resulted sub-matrices are stored in a numpy array of shape (X, side, side, 1).

write_sparse_matrix(matrix_type, path)[source]¶

The reconstructed and predicted Hi-C matrix is saved in a sparse matrix file.

Parameters:	matrix_type (str) – Matrix’s name path (str) – Path of the output

Interpolation¶

class src.interpolation.Interpolation(alphas)[source]¶

class Interpolation¶
This class groups attributes and functions which aim to construct, write in a sparse matrix
and plot two or several interpolated matrices.

alphas¶

List of float values to use for the interpolation (alpha parameter)

Type:	list

interpolated_submatrices¶

List of all the interpolated sub-matrices. Each item in the list contains an interpolation with a different alpha.

Type:	list

integrated_matrix¶

List of all the integrated (interpolated) reconstructed matrices. Each item in the list contains an interpolation with a different alpha.

Type:	list

construct_integrated_matrix(hic)[source]¶

Construction of the whole integrated matrices from the interpolated sub-matrices.

Parameters:	hic (Hic(Matrix) object) – Hi-C matrix

plot_integrated_matrix(hic, color_map, path)[source]¶

The integrated matrices are plotted for each alpha value.

Parameters:	hic (Hic(Matrix) object) – Hi-C matrix color_map (matplotlib.colors.ListedColormap) – Color map path (str) – Path of the output plot

plot_interpolated_submatrices(hic, index_list, color_map, path)[source]¶

40 random integrated sub-matrices are plotted for each alpha value.

Parameters:	hic (Hic(Matrix) object) – Hi-C matrix index_list (list) – List of the 40 sub-matrix indexes to plot color_map (matplotlib.colors.ListedColormap) – Color map path (str) – Path of the output plot

write_predicted_sparse_matrix(hic, path, threshold=0.0001)[source]¶

The integrated matrices are saved in sparse matrix files for each alpha value.

Parameters:	hic (Hic(Matrix) object) – Hi-C matrix path (str) – Path of the output threshold (float) – The values under the threshold will be set to 0

class src.interpolation.InterpolationInLatentSpace(*args, **kwargs)[source]¶

class InterpolationInLatentSpace¶
This class inherits the Interpolation class and interpolate sub-matrices in the latent space

interpolate_latent_spaces(hist_marks, hic_latent_spaces)[source]¶

Double linear interpolation of the latent spaces of the Hi-C and histone marks.

Parameters:	hist_marks (dict) – Dictionary containing all histone mark HistoneMark objects. predicted_hic (numpy array) – Predicted sub-matrices of the Hi-C

set_decoded_latent_spaces(decoder, side)[source]¶

The interpolated latent spaces are decoded.

Parameters:	decoder (keras model object) – Hi-C matrix side (int) – Square side

class src.interpolation.NormalInterpolation(*args, **kwargs)[source]¶

class InterpolationInLatentSpace¶
This class inherits the Interpolation class and interpolate sub-matrices in the pixel space
(= without the use of encoder and decoder).