Custom transformation API (variant of scanpy.pp.normalize_total such as scanpy.pp.normalize)

• Additional function parameters / changed functionality / changed defaults?
• New analysis tool: A simple analysis tool you have been using and are missing in sc.tools?
• New plotting function: A kind of plot you would like to seein sc.pl?
• External tools: Do you know an existing package that should go into sc.external.*?
• Other?

...

I'm in the process of redesigning my most used workflows from https://github.com/jolespin/soothsayer to wrap around ScanPy. I come from microbial ecology but do a good amount of machine learning. In microbial ecology, the community is shifting towards a compositional data analysis (CoDA) approach which has fundamentals firmly rooted in mathematics.

Here is some literature about broad-scale applications across all NGS datasets:

• A field guide for the compositional analysis of any-omics data
• Understanding sequencing data as compositions: an outlook and review

it is even catching attention in scRNA-seq too:

• scCODA is a Bayesian model for compositional single-cell data analysis

Anyways, off my soap box. As mentioned, I'm in the process of adapting my workflows to take advantage of ScanPy's power but I'm having a few difficulties. The first incorporating custom transformations.

In future versions, would it be possible to create an API that is similar to [scanpy.pp.normalize_total)(https://scanpy.readthedocs.io/en/stable/generated/scanpy.pp.normalize_total.html) but allows for a custom metric?

For example:

import numpy as np
import pandas as pd
from typing import Union

def clr(x:Union[np.ndarray, pd.Series], multiplicative_replacement:Union[None,str,float,int]="auto") -> Union[np.ndarray, pd.Series]:
    """
    http://scikit-bio.org/docs/latest/generated/skbio.stats.composition.clr.html#skbio.stats.composition.clr
    """
    assert np.all(x >= 0)
    if multiplicative_replacement == "auto":
        if np.any(x == 0):
            multiplicative_replacement = 1/(len(x)**2)
    if multiplicative_replacement is None:
        multiplicative_replacement = 0
    x = x.copy() + multiplicative_replacement
    x = x/x.sum()
    log_x = np.log(x)
    geometric_mean = log_x.mean()
    return log_x - geometric_mean

kwargs = {"multiplicative_replacement":"auto"}

sc.pp.normalize(adata, function=clr, **kwargs) # -> addata object with the transformation as one of the layers

Start with the scanpy.pp.normalize_total API and the issue's clr example to understand the requested custom transformation entry point. Done would be a decided and documented API that accepts a custom metric with keyword arguments and produces the transformed data in a layer, as described.