An error happend with “palantir.preprocess.log_transform ” function, and with error message "Adding a nonzero scalar to a sparse matrix is not supported

[Infertility_Knowledge]

 I tried to follow the example of "Palantir analysis notebook", and an error message was reported with “palantir.preprocess.log_transform ” function,  and an error message with "Adding a nonzero scalar to a sparse matrix is not supported".
The original code  with  

import palantir
import scanpy as sc
import numpy as np
import pandas as pd
import os

Plotting

import matplotlib
import matplotlib.pyplot as plt
import seaborn as sns
import warnings

Inline plotting

%matplotlib inline

sns.set_style('ticks')
matplotlib.rcParams['figure.figsize'] = [4, 4]
matplotlib.rcParams['figure.dpi'] = 100
matplotlib.rcParams['image.cmap'] = 'Spectral_r'
warnings.filterwarnings(action="ignore", module="matplotlib", message="findfont")

Reset random seed

np.random.seed(5)

Load sample data

palantir_dir = os.path.expanduser('data/')
ad = sc.read(palantir_dir + 'marrow_sample_scseq_counts.h5ad')
ad
sc.pp.normalize_per_cell(ad)
palantir.preprocess.log_transform(ad)
Error message

NotImplementedError Traceback (most recent call last)
Input In [12], in <cell line: 1>()
----> 1 palantir.preprocess.log_transform(ad)

File C:\ProgramData\Anaconda3\envs\r-reticulate\lib\site-packages\palantir\preprocess.py:43, in log_transform(data, pseudo_count)
37 """Log transform the matrix
38
39 :param data: Counts matrix: Cells x Genes
40 :return: Log transformed matrix
41 """
42 if type(data) is sc.AnnData:
---> 43 data.X = np.log2(data.X + pseudo_count) - np.log2(pseudo_count)
44 else:
45 return np.log2(data + pseudo_count)

File C:\ProgramData\Anaconda3\envs\r-reticulate\lib\site-packages\scipy\sparse\base.py:410, in spmatrix.add(self, other)
408 return self.copy()
409 # Now we would add this scalar to every element.
--> 410 raise NotImplementedError('adding a nonzero scalar to a '
411 'sparse matrix is not supported')
412 elif isspmatrix(other):
413 if other.shape != self.shape:

NotImplementedError: adding a nonzero scalar to a sparse matrix is not supported
palantir/SciPy/NumPy/Python version information
1.0.1 1.9.0 1.22.3 sys.version_info(major=3, minor=9, micro=12, releaselevel='final', serial=0)

[Manu Setty]

Hello - very sorry for the delayed response.

You should be able to resolve this issues using

from spicy.sparse import csr_matrix
ad.X = csr_matrix(ad.X)

[Maren Büttner]

Hi @ManuSetty

I tried the same solution, but it does not seem to solve the issue described above. I took the liberty to inspect the log_transform function in palantir. My suggestion is to replace if type(ad) is sc.Anndata by if isinstance(ad, anndata.AnnData) and import anndata in your header.

I tested the adapted log_transform function in Google Colabs and it worked fine.

[Manu Setty]

Thank you!