DEM: error when foreground and background regions have the same number of samples

[SeppeDeWinter]

When a contrast is given where fore and background have the same number of regions the following error occurs:

File /opt/venv/lib/python3.8/site-packages/pycistarget/motif_enrichment_dem.py:706, in DEM_internal(***failed resolving arguments***)
    704 keep_motifs = motif_df.index.tolist()
    705 keep_motifs_index = get_position_index(keep_motifs, motifs)
--> 706 scores_mat = sparse.vstack([fg_mat[keep_motifs_index,].T, bg_mat[keep_motifs_index,].T], format='csr').T
    707 regions = region_group[0] + ['Bg']*bg_mat.shape[1]
    708 labels = [1]*fg_mat.shape[1] + [0]*bg_mat.shape[1]

File /opt/venv/lib/python3.8/site-packages/scipy/sparse/_construct.py:553, in vstack(blocks, format, dtype)
    522 def vstack(blocks, format=None, dtype=None):
    523     """
    524     Stack sparse matrices vertically (row wise)
    525
   (...)
    551
    552     """
--> 553     return bmat([[b] for b in blocks], format=format, dtype=dtype)

File /opt/venv/lib/python3.8/site-packages/scipy/sparse/_construct.py:602, in bmat(blocks, format, dtype)
    599 blocks = np.asarray(blocks, dtype='object')
    601 if blocks.ndim != 2:
--> 602     raise ValueError('blocks must be 2-D')
    604 M,N = blocks.shape
    606 # check for fast path cases

ValueError: blocks must be 2-D

this occurs because internally in the sparse.vstack function the following is run.

np.array([fg_mat[keep_motifs_index,].T, bg_mat[keep_motifs_index,].T], object)

if both fg_mat and bg_mat have the same number of regions this will result in a np.array with dimension (2, n_regions, n_motfs) (this causes the error).

Ff they have different number of regions this will result in a np.array of two np.arrays.