hic3defdr.util.matrices module

hic3defdr.util.matrices.deconvolute(matrix, bias, invert=False)

Applies bias factors to a sparse matrix.

Parameters:

• matrix (scipy.sparse.spmatrix) – The matrix to bias. Will be converted to CSR by this function.
• bias (np.ndarray) – The dense bias vector.
• invert (bool) – Whether or not to invert the bias before applying it. By default this function multiplies the matrix by the bias. Infinite bias (1 / 0) will be converted to zero bias to allow rows with infinite bias to be dropped from the resulting sparse matrix.

Returns:

The deconvoluted matrix.

Return type:

scipy.sparse.csr_matrix

hic3defdr.util.matrices.dilate(matrix, doublings)

Doubles the “resolution” of a matrix using nearest-neighbor interpolation.

Parameters:

• matrix (np.ndarray) – Input matrix.
• doublings (int) – The number of times the resolution will be doubled.

Returns:

The dilated matrix.

Return type:

np.ndarray

hic3defdr.util.matrices.select_matrix(row_slice, col_slice, row, col, data, symmetrize=True)

Slices out a dense matrix from COO-formatted sparse data, filling empty values with np.nan.

Parameters:

• col_slice (row_slice,) – Row and column slice to use, respectively.
• col, data (row,) – COO-format sparse matrix to be sliced.
• symmetrize (bool) – Pass True to fill in lower-triangle points of the matrix.

Returns:

Dense matrix.

Return type:

np.ndarray

hic3defdr.util.matrices.sparse_intersection(fnames, bias=None)

Computes the intersection set of (row, col) pairs across multiple sparse matrices.

Parameters:

• fnames (list of str) – File paths to sparse matrices loadable by scipy.sparse.load_npz(). Will be converted to COO by this function.
• bias (np.ndarray, optional) – Pass the bias matrix to drop rows and columns with bias factors of zero in any replicate.

Returns:

row, col – The intersection set of (row, col) pairs.

Return type:

np.ndarray

hic3defdr.util.matrices.sparse_union(fnames, dist_thresh=1000, bias=None, size_factors=None, mean_thresh=0.0)

Computes the intersection set of (row, col) pairs across multiple sparse matrices.

Parameters:

• fnames (list of str) – File paths to sparse matrices loadable by scipy.sparse.load_npz(). Will be converted to COO by this function.
• dist_thresh (int) – The maximum distance allowed, respectively, in bin units.
• bias (np.ndarray, optional) – Rectangular matrix containing bias factors for each bin (rows) and each replicate (columns).
• size_factors (np.ndarray, optional) – Size factors for each replicate.
• mean_thresh (float) – Minimum mean value (in normalized space) to keep pixels for.

Returns:

row, col – The union set of (row, col) pairs.

Return type:

np.ndarray

hic3defdr.util.matrices.wipe_distances(matrix, min_dist, max_dist)

Eliminates entries from a sparse matrix outside of a specified distance range.

Parameters:

• matrix (scipy.sparse.spmatrix) – The matrix to wipe. Will be converted to COO by this function.
• max_dist (min_dist,) – The minimum and maximum distance allowed, respectively, in bin units.

Returns:

The wiped matrix.

Return type:

scipy.sparse.coo_matrix