Additional parameters in some functions - new functions

Dear all,

I am writing to ask you some other functionalities. I have just moved from Seurat to Scanpy and I am finding Scanpy a very nice and well done Python package.

1. I wrote a function to show the 3D plot of the UMAP, tSNE and PCA spaces. In the scanpy.tl.tsne function is not possible to change the number of components, it calculates only the first two components, even if the scanpy.pl.tsne function has a parameter component. May you add a parameter like the n_components of the scanpy.tl.umap function?

2. In the rank_genes_groups function the log2FC values are provided only for ‘t-test’ based methods. May you return the log2FC values (maybe named log2FC) for all the implemented statistical methods?

3. I think that two parameters in the rank_genes_groups function should be added.

   • min_pCells to test only the genes that are detected in a minimum fraction of cells of either of the two populations (e.g., cluster 0 vs rest). For instance, min_pCells=0.3 means that at least 30% of the cells must express that gene.
   • positive, if it is True, the function should return only positive marker genes for each population.

4. A function showing the volcano plots (based on the log2FC) can help (I can write it if the log2FC values are provided).

Thank you in advance. Best, Andrea

[ivirshup]

Sorry this is a little off topic, but it’s something I’ve found useful:

@LuckyMD and anyone else looking for an interactive volcano plot, I’ve been using hvplot in my notebooks. A volcano plot can be made from DE data frame with something like:

de_df.hvplot.scatter(
    "logfoldchanges", "pvals_adj", 
    flip_yaxis=True, logy=True, 
    hover_cols=["names"]
)

Complete example using scanpy

import pandas as pd
import numpy as np
import hvplot.pandas
import scanpy as sc

def rank_genes_groups_df(adata, group, pval_cutoff : float =None, logfc_cutoff=None): 
    d = pd.DataFrame() 
    for k in ['scores', 'names', 'logfoldchanges', 'pvals', 'pvals_adj']: 
        d[k] = adata.uns["rank_genes_groups"][k][group] 
    if pval_cutoff is not None: 
        d = d[d["pvals_adj"] < pval_cutoff] 
    if logfc_cutoff is not None: 
        d = d[d["logfoldchanges"].abs() > logfc_cutoff] 
    return d

pbmcs = sc.datasets.pbmc68k_reduced()
sc.tl.rank_genes_groups(pbmcs, "bulk_labels", n_genes=pbmcs.var_names.size)
de_df = rank_genes_groups_df(pbmcs, "CD34+")

de_df.hvplot.scatter(
    "logfoldchanges", "pvals_adj", 
    flip_yaxis=True, logy=True, 
    hover_cols=["names"]
)

[andrea-tango]

@LuckyMD your tutorial is very interesting! I agree with you regarding the min_pCells parameter, it should be used as a filtering step before calculating the marker genes.