Mouse Aging 10X RNA-seq gene expression

In this notebook we’ll explore some gene expressions and combine them with the cell metadata we showed in the previous clustering analysis tutorial.

You need to be either connected to the internet or connected to a cache that has the aging data downloaded already and have already downloaded the example data via the getting started notebook to run this notebook.

import pandas as pd
import numpy as np
import anndata
from pathlib import Path
import matplotlib.pyplot as plt
import matplotlib.colors as colors
from typing import Optional, Tuple

from abc_atlas_access.abc_atlas_cache.abc_project_cache import AbcProjectCache

We will interact with the data using the AbcProjectCache. This cache object tracks which data has been downloaded and serves the path to the requested data on disk. For metadata, the cache can also directly serve up a Pandas Dataframe. See the getting_started notebook for more details on using the cache including installing it if it has not already been.

Change the download_base variable to where you have downloaded the data in your system.

download_base = Path('../../data/abc_atlas')
abc_cache = AbcProjectCache.from_cache_dir(download_base)

abc_cache.current_manifest

'releases/20241130/manifest.json'

Create the expanded cell metadata as was done previously in the cluster annotation tutorial.

# Load the cell metadata.
cell = abc_cache.get_metadata_dataframe(
    directory='Zeng-Aging-Mouse-10Xv3',
    file_name='cell_metadata',
    dtype={'cell_label': str,
           'wmb_cluster_alias': 'Int64'}
)
cell.set_index('cell_label', inplace=True)

cell_colors = abc_cache.get_metadata_dataframe(
    directory='Zeng-Aging-Mouse-10Xv3',
    file_name='cell_annotation_colors'
).set_index('cell_label')

cluster_info = abc_cache.get_metadata_dataframe(
    directory='Zeng-Aging-Mouse-10Xv3',
    file_name='cluster'
).set_index('cluster_alias')

cell_cluster_mapping = abc_cache.get_metadata_dataframe(
    directory='Zeng-Aging-Mouse-WMB-taxonomy',
    file_name='cell_cluster_mapping_annotations'
).set_index('cell_label')
cell_cluster_mapping.head()

# Join on the cell_label index.
cell_extended = cell.join(cell_cluster_mapping, rsuffix='_cl_map')
cell_extended = cell_extended.join(cell_colors, rsuffix='_cl_colors')
# Join the cluster information in by joining on the Aging dataset's cluster_alias column.
cell_extended = cell_extended.join(cluster_info, on='cluster_alias', rsuffix='_cl_info')

# Quick run through to drop duplicated columns
drop_cols = []
for col in cell_extended.columns:
    if col.endswith(('_cl_map', '_cl_colors', '_cl_info')):
        drop_cols.append(col)
cell_extended.drop(columns=drop_cols, inplace=True)

# The dataset is sorted on cell_label by default, this causes some plotting weirdness
# due to all "adult" cells being first in the order. Below we scrabble the DataFrame
# to better reproduce plots from the paper.
cell_extended = cell_extended.sample(frac=1, random_state=12345)

del cell
del cell_colors
del cell_cluster_mapping
del cluster_info

cell_extended.head()

	cell_barcode	gene_count	umi_count	doublet_score	x	y	cluster_alias	cell_in_wmb_study	wmb_cluster_alias	library_label	...	proportion_adult_cells	proportion_aged_cells	odds_ratio	log2_odds_ratio	cluster_age_bias	max_region_of_interest_color	cluster_age_bias_color	neurotransmitter_combined_label	neurotransmitter_label	neurotransmitter_color
cell_label
GCCTGTTGTGAATTAG-135_B01	GCCTGTTGTGAATTAG	6777	37834.0	0.200000	-1.185296	1.881103	278	True	1079	L8TX_190716_01_D07	...	0.577949	0.422051	1.013716	0.019653	unassigned	#80CDF8	#DADEDF	GABA	GABA	#FF3358
TCCGAAAGTGAAGCGT-761_A04	TCCGAAAGTGAAGCGT	3396	9585.0	0.030303	14.007499	-0.220336	817	False	<NA>	L8TX_210805_01_H01	...	0.454494	0.545506	1.046448	0.065500	unassigned	#8599CC	#DADEDF	No-NT	No-NT	#666666
TTACCATGTCGTGGTC-327_A06	TTACCATGTCGTGGTC	4294	10527.0	0.020000	6.550781	-1.378886	804	False	<NA>	L8TX_200813_01_H10	...	0.370570	0.629430	1.741208	0.800089	unassigned	#80C0E2	#DADEDF	No-NT	No-NT	#666666
ACGGGTCGTACGAGCA-385_D06	ACGGGTCGTACGAGCA	2943	6957.0	0.000000	13.650183	0.289086	817	False	<NA>	L8TX_201008_01_A12	...	0.454494	0.545506	1.046448	0.065500	unassigned	#8599CC	#DADEDF	No-NT	No-NT	#666666
CTTAGGATCTGTCCCA-301_B04	CTTAGGATCTGTCCCA	7907	47406.0	0.037037	6.583878	-10.639808	152	False	<NA>	L8TX_200723_01_B10	...	0.403409	0.596591	0.587232	-0.767999	unassigned	#72D569	#DADEDF	Glut	Glut	#2B93DF

5 rows × 60 columns

Single cell transcriptomes

The ~1.2 million cell dataset of Aging Mouse is contained in one gene expression matrix for each of the raw and log2 normalized data. Each matrix file is formatted as an annadata, h5ad file with minimal metadata.

Below we show some interactions with data from the 10X expression matrices in the Aging Mouse dataset. For a deeper dive into how to access specific gene data from the expression matrices, se the general_acessing_10x_snRNASeq_tutorial.ipynb notebook. Below we will use precomputed metadata from these matrices to look at gene expression both in relation to different neurotransmitters and locations across the brain.

First, we list the available metadata in the Zeng-Aging-Mouse-10Xv3 dataset again. The file we will be using in this part of the tutorial is the example_genes_all_cells_expression table. We will load the genes from the WMB-10X dataset as the genes for the aging data are identical.

The table below holds metadata for all genes sequenced in the dataset.

gene = abc_cache.get_metadata_dataframe(directory='WMB-10X', file_name='gene').set_index('gene_identifier')
print("Number of genes = ", len(gene))
gene.head(5)

Number of genes =  32285

	gene_symbol	name	mapped_ncbi_identifier	comment
gene_identifier
ENSMUSG00000051951	Xkr4	X-linked Kx blood group related 4	NCBIGene:497097	NaN
ENSMUSG00000089699	Gm1992	predicted gene 1992	NaN	NaN
ENSMUSG00000102331	Gm19938	predicted gene, 19938	NaN	NaN
ENSMUSG00000102343	Gm37381	predicted gene, 37381	NaN	NaN
ENSMUSG00000025900	Rp1	retinitis pigmentosa 1 (human)	NCBIGene:19888	NaN

We’ll skip accessing these data from the expression matrices specifically for now, however, users can learn how to access specific genes from the released expression matrices in thegeneral_acessing_10x_snRNASeq_tutorial.ipynb notebook.

The precomputed table below contains expressions for the genes Slc17a7, Slc17a6, Slc17a8, Slc32a1, Slc6a5, Slc6a3, Slc6a4, and ‘Tac2 for all cells across the Aging Mouse dataset. We then join this with our previously created cell_extended pandas DataFrame from this tutorial.

gene_names = ['Slc17a7', 'Slc17a6', 'Slc17a8', 'Slc32a1', 'Slc6a5', 'Slc6a3', 'Slc6a4', 'Tac2']
example_genes = abc_cache.get_metadata_dataframe(
    directory='Zeng-Aging-Mouse-10Xv3',
    file_name='example_genes_all_cells_expression'
).set_index('cell_label')

cell_extended_with_genes = cell_extended.join(example_genes)
cell_extended_with_genes.head()

	cell_barcode	gene_count	umi_count	doublet_score	x	y	cluster_alias	cell_in_wmb_study	wmb_cluster_alias	library_label	...	neurotransmitter_label	neurotransmitter_color	Slc32a1	Slc17a7	Slc6a5	Slc17a6	Slc17a8	Tac2	Slc6a4	Slc6a3
cell_label
GCCTGTTGTGAATTAG-135_B01	GCCTGTTGTGAATTAG	6777	37834.0	0.200000	-1.185296	1.881103	278	True	1079	L8TX_190716_01_D07	...	GABA	#FF3358	7.900150	0.000000	0.0	0.000000	0.0	0.0	0.0	0.0
TCCGAAAGTGAAGCGT-761_A04	TCCGAAAGTGAAGCGT	3396	9585.0	0.030303	14.007499	-0.220336	817	False	<NA>	L8TX_210805_01_H01	...	No-NT	#666666	0.000000	0.000000	0.0	0.000000	0.0	0.0	0.0	0.0
TTACCATGTCGTGGTC-327_A06	TTACCATGTCGTGGTC	4294	10527.0	0.020000	6.550781	-1.378886	804	False	<NA>	L8TX_200813_01_H10	...	No-NT	#666666	0.000000	0.000000	0.0	0.000000	0.0	0.0	0.0	0.0
ACGGGTCGTACGAGCA-385_D06	ACGGGTCGTACGAGCA	2943	6957.0	0.000000	13.650183	0.289086	817	False	<NA>	L8TX_201008_01_A12	...	No-NT	#666666	0.000000	0.000000	0.0	0.000000	0.0	0.0	0.0	0.0
CTTAGGATCTGTCCCA-301_B04	CTTAGGATCTGTCCCA	7907	47406.0	0.037037	6.583878	-10.639808	152	False	<NA>	L8TX_200723_01_B10	...	Glut	#2B93DF	4.465607	8.794357	0.0	7.864422	0.0	0.0	0.0	0.0

5 rows × 68 columns

Example use cases

During analysis, clusters were assigned neurotransmitter identities based on the expression of canonical neurotransmitter transporter genes. In this example, we create a DataFrame comprising of expression of the 9 solute carrier family genes for all the cells in the dataset. We then group the cells by the assigned neurotransmitter class and compute the mean expression for each group. We will define a simple helper function to visualize the expression as a heatmap.

Below we define a helper function aggregate_by_metadata to compute the average expression for a given category and later plot_umap to plot cells in a UMAP colorized by metadata or expression values like what was used in the previous tutorial.

def aggregate_by_metadata(df, gnames, value, sort = False):
    grouped = df.groupby(value)[gnames].mean()
    if sort:
        grouped = grouped.sort_values(by=gnames[0], ascending=False)
    return grouped

Expression of selected genes in the whole brain

The helper function below creates a heatmap showing the relation between various parameters in the combined cell metadata and the genes we loaded. For the default colormap used, darker colors equate to a higher expression.

def plot_heatmap(df, fig_width=8, fig_height=4, cmap=plt.cm.magma_r, vmin=None, vmax=None) :

    arr = df.to_numpy().astype('float')

    fig, ax = plt.subplots()
    fig.set_size_inches(fig_width, fig_height)

    im = ax.imshow(arr, cmap=cmap, aspect='auto', vmin=vmin, vmax=vmax)
    xlabs = df.columns.values
    ylabs = df.index.values

    ax.set_xticks(range(len(xlabs)))
    ax.set_xticklabels(xlabs)

    ax.set_yticks(range(len(ylabs)))
    res = ax.set_yticklabels(ylabs)
    
    return im

Below, we plot the expression of the genes we’ve selected for this example. We show the genes vs neurotransmitter type/identity. As expected, the genes associated with the various neurotransmitter express as expected.

agg = aggregate_by_metadata(cell_extended_with_genes, gene_names, 'neurotransmitter_label')
res = plot_heatmap(agg, 8, 8)
plt.show()

By dissection region of interest shows that each of these genes are associated with distinct regions.

agg = aggregate_by_metadata(cell_extended_with_genes, gene_names, 'anatomical_division_label')
res = plot_heatmap(agg, 8, 8)
plt.show()

And finally, by class from the taxonomy.

agg = aggregate_by_metadata(cell_extended_with_genes, gene_names, 'class_name')
res = plot_heatmap(agg, 8, 12)
plt.show()

We can also visualize the relationship between these genes and their location in the UMAP.

def plot_umap(xx, yy, cc=None, val=None, fig_width=8, fig_height=8, cmap=None):

    fig, ax = plt.subplots()
    fig.set_size_inches(fig_width, fig_height)

    if cmap is not None :
        plt.scatter(xx, yy, s=0.5, c=val, marker='.', cmap=cmap)
    elif cc is not None :
        plt.scatter(xx, yy, s=0.5, color=cc, marker='.')
        
    ax.axis('equal')
    # ax.set_xlim(-18, 27)
    # ax.set_ylim(-18, 27)
    ax.set_xticks([])
    ax.set_yticks([])
    
    return fig, ax

for gene_name in gene_names:
    fig, ax = plot_umap(
        cell_extended_with_genes['x'],
        cell_extended_with_genes['y'],
        val=cell_extended_with_genes[gene_name],
        cmap=plt.cm.magma_r)
    res = ax.set_title(f"Gene {gene_name}")
    plt.show()
    break # Remove break statement to see all selected genes

AgeDE Genes

The Aging Mouse data also provides a set of aging Differential Expression (ageDE) genes at various levels of the taxonomy. The table is presented as a flattened set of ageDE genes found within a specific cell type where each row is a unique gene/celltype combination. Note that for a given level in the taxonomy, only certain cell types have significant ageDE genes reported. For instance, the supertype level in the aging_degenes table only contains non-neuronal cell types.

Below we load and display this table.

degenes = abc_cache.get_metadata_dataframe(
    directory='Zeng-Aging-Mouse-WMB-taxonomy',
    file_name='aging_degenes'
)
degenes.head()

	grouping_type	grouping_label	grouping_name	gene_identifier	gene_symbol	is_primary_ieg	age_effect_size	unadjusted_pvalue	adjusted_pvalue	confidence_interval_higher_bound	confidence_interval_lower_bound
0	subclass	CS20230722_SUBC_100	100 AHN Onecut3 Gaba	ENSMUSG00000073590	3222401L13Rik	False	1.812935	7.766049e-12	9.653199e-08	2.346020	1.279850
1	subclass	CS20230722_SUBC_100	100 AHN Onecut3 Gaba	ENSMUSG00000095041	AC149090.1	False	1.289552	3.322308e-17	4.129629e-13	1.793447	0.785656
2	subclass	CS20230722_SUBC_100	100 AHN Onecut3 Gaba	ENSMUSG00000086503	Xist	False	1.185049	2.222176e-11	2.762165e-07	1.645855	0.724244
3	subclass	CS20230722_SUBC_132	132 AHN-RCH-LHA Otp Fezf1 Glut	ENSMUSG00000073590	3222401L13Rik	False	1.867499	1.532389e-15	1.963297e-11	2.366093	1.368905
4	subclass	CS20230722_SUBC_132	132 AHN-RCH-LHA Otp Fezf1 Glut	ENSMUSG00000095041	AC149090.1	False	1.229435	4.989205e-25	6.392170e-21	1.589657	0.869213

Additionally, we load a mapping of all clusters to the levels higher in the taxonomy complete with their colors.

# mapping = pd.read_csv(taxonomy_base / 'cluster_mapping_pivot.csv' )
mapping = abc_cache.get_metadata_dataframe(
    directory='Zeng-Aging-Mouse-WMB-taxonomy',
    file_name='cluster_mapping_pivot'
)
mapping.head()

	cluster_alias	cluster_label	cluster_order	cluster_name	cluster_color	class_name	subclass_name	supertype_name	class_color	subclass_color	supertype_color
0	1	CS20241021_0001	0	1_CLA-EPd-CTX Car3 Glut_1	#f7c27a	01 IT-ET Glut	001 CLA-EPd-CTX Car3 Glut	0001 CLA-EPd-CTX Car3 Glut_1	#FA0087	#64c2fc	#99822E
1	2	CS20241021_0002	1	2_IT EP-CLA Glut_1	#5c1a34	01 IT-ET Glut	002 IT EP-CLA Glut	0003 IT EP-CLA Glut_1	#FA0087	#1F665D	#994563
2	3	CS20241021_0003	2	3_IT EP-CLA Glut_1	#b96cf0	01 IT-ET Glut	002 IT EP-CLA Glut	0003 IT EP-CLA Glut_1	#FA0087	#1F665D	#994563
3	4	CS20241021_0004	3	4_IT EP-CLA Glut_1	#bff691	01 IT-ET Glut	002 IT EP-CLA Glut	0003 IT EP-CLA Glut_1	#FA0087	#1F665D	#994563
4	5	CS20241021_0005	4	5_IT EP-CLA Glut_1	#39fbfa	01 IT-ET Glut	002 IT EP-CLA Glut	0003 IT EP-CLA Glut_1	#FA0087	#1F665D	#994563

ageDE vs supertype/subclass/cluster

We declare a convenience function for working with our two tables above. This calculates the number of ageDE genes for a selected level in the taxonomy (cluster, supertype, subclass) and plots them as a bar graph.

def plot_de_gene_count_by_taxonomy(
        mapping_df: pd.DataFrame,
        degene_df: pd.DataFrame,
        level_name: str,
        color_name: str,
        title: Optional[str]=None,
        classes: Optional[list[str]]=None,
        xlim: Tuple[int, int]=(0, 400),
        figsize: Tuple[int, int]=(3.5, 8)
) -> pd.DataFrame:
    """Plot ageDE genes grouped by cluser, supertype, subclass.

    Parameters
    ----------
    mapping_df: pandas.DataFrame
        DataFrame mapping of cluster to higher taxonomy levels with colors.
    degene_df: pandas.DataFrame
        DataFrame of ageDE genes with their associated cell type.
    level_name: str
        Name of taxonomy level to plot.
    color_name: str
        The name of the column to color by.
    title: str
        Title to add to the plot
    classes: list of str
        Plot only selected classes instead of all data.
    xlim: tuple of ints
        Plot limit on the x axis.
    figsize: tuple of float
        Plot size in inches.
    """
    # Get our colors
    color_map = mapping_df[[level_name, color_name, 'class_name']].copy()
    color_map.drop_duplicates(inplace=True)
    color_map.set_index(level_name, inplace=True)

    # Get counts for level (e.g. supertype_name)
    grouped = degene_df.groupby(['grouping_type', 'grouping_name'])['gene_symbol'].count()[level_name.split('_')[0]]

    # Setup our figure
    fig, ax = plt.subplots(1, 1)
    fig.set_size_inches(figsize[0], figsize[1])

    class_map = []
    for name in grouped.index.values:
        submapping = mapping[mapping[level_name] == name].iloc[0]
        class_map.append({
            "grouping_name": name,
            "class_name": submapping['class_name'],
        })
    class_map = pd.DataFrame(data=class_map).set_index("grouping_name")

    if classes:
        class_mask = class_map[class_map['class_name'].isin(classes)].index.values
    else:
        class_mask = grouped.index.values
    
    counts = grouped.loc[class_mask].values[::-1]
    names = grouped.loc[class_mask].index.values[::-1]
    colors = color_map.loc[names, color_name].values

    # plot
    ax.barh(names, counts, color=colors)
    ax.set_xlim([xlim[0], xlim[1]])
    # ax.set_ylim([0, max_n_bars])
    ax.set_title(title)
    plt.show()

vs Non-neuronal supertype

We’ll use groupby over supertype to plot the non-neuronal ageDE genes colored by class.

plot_de_gene_count_by_taxonomy(
    mapping_df=mapping,
    degene_df=degenes,
    level_name='supertype_name',
    color_name='class_color'
)

vs Neuronal subclasses

And now the same but this time for two of the subclasses. We plot only subclasses from classes, 12 HY GABA and 14 HY Glut for readablity of the plot. Add or select other classes to plot those as well.

plot_de_gene_count_by_taxonomy(
    mapping_df=mapping,
    degene_df=degenes,
    level_name='subclass_name',
    color_name='class_color',
    figsize=(3.5, 8),
    classes=['12 HY GABA', '14 HY Glut']
)

vs Neuronal clusters

And now the same but this time for the neuronal clusters. We plot only clusters from the class 12 HY GABA again for readablity of the plot.

plot_de_gene_count_by_taxonomy(
    mapping_df=mapping,
    degene_df=degenes,
    level_name='cluster_name',
    color_name='class_color',
    figsize=(3.5, 8),
    classes=['12 HY GABA']
)

Top N ageDE genes.

We’ll write a quick function to extract the top N genes by p-value for a given supertype, subclass etc. and display the new, smaller ageDE gene table.

def top_ageDE_genes(
    degene_df: pd.DataFrame,
    level: Optional[str]=None,
    top_n: int=3
):
    if level is None:
        level_degenes = degene_df
    else:
        level_degenes = degene_df[degene_df['grouping_type'] == level]

    output_df = []
    celltypes = level_degenes['grouping_name'].unique()
    for celltype in celltypes:
        subdf = degenes[degenes['grouping_name'] == celltype].sort_values(by='adjusted_pvalue')[::-1]
        if len(subdf) < top_n:
            output_df.append(subdf)
        else:
            output_df.append(subdf[:top_n])
    return pd.concat(output_df).reset_index(drop=True)

Below we’ll grab the top 5 genes for all the supertypes. We’ll show only the first 5 rows of the new table. This shows the top 5 genes by adjusted p_value for 1186 ABC NN_1. This analysis can be repeated for any top N genes and any of the celltype taxonomy levels.

top_ageDE = top_ageDE_genes(degene_df=degenes, level='supertype', top_n=5)
top_ageDE.head()

	grouping_type	grouping_label	grouping_name	gene_identifier	gene_symbol	is_primary_ieg	age_effect_size	unadjusted_pvalue	adjusted_pvalue	confidence_interval_higher_bound	confidence_interval_lower_bound
0	supertype	CS20230722_SUPT_1186	1186 ABC NN_1	ENSMUSG00000002602	Axl	False	1.443417	8.056410e-07	0.008445	2.006174	0.880659
1	supertype	CS20230722_SUPT_1186	1186 ABC NN_1	ENSMUSG00000019929	Dcn	False	-1.809868	6.742604e-07	0.007068	-1.104921	-2.514814
2	supertype	CS20230722_SUPT_1186	1186 ABC NN_1	ENSMUSG00000049148	Plcxd3	False	1.054319	4.960283e-07	0.005199	1.571687	0.536952
3	supertype	CS20230722_SUPT_1186	1186 ABC NN_1	ENSMUSG00000079547	H2-DMb1	False	1.707299	4.853269e-07	0.005087	2.326934	1.087663
4	supertype	CS20230722_SUPT_1186	1186 ABC NN_1	ENSMUSG00000052974	Cyp2f2	False	-1.323712	2.652985e-07	0.002781	-0.836481	-1.810943

Select ageDE expression vs subclass

Next we’ll plot the expression of a select set of ageDE genes vs the subclasses in the taxonomy. The genes we select here are a subset of the those plotted in Extended Figure 7 from [Jin et al.], however one could plot all ageDE genes by modifying the degene_list. Note also that the ordering of the subclasses is different then the ordering in that plot.

First we’ll declare a function to extract the specific genes and celltype level we would like to plot.

def get_genes_for_celltype_level(
    degene_df: pd.DataFrame,
    level: str,
    degene_list: list[str],
):
    level_degenes = degene_df[degene_df['grouping_type'] == level]
    levels = sorted(level_degenes['grouping_name'].unique())
    level_by_gene = pd.DataFrame(columns=degene_list, index=levels)
    for idx, derow in level_degenes.iterrows():
        if derow['gene_symbol'] in degene_list:
            level_by_gene.loc[derow['grouping_name'], derow['gene_symbol']] = derow['age_effect_size']
    return level_by_gene

Let’s first plot a select set of genes for the subclasses in the ageDE gene expression table. Note that that you can modify the list of genes to plot any subset or the full set of genes in the ageDE expression table by specifying the gene symbol.

subclass_by_gene = get_genes_for_celltype_level(
    degene_df=degenes,
    level='subclass',
    degene_list=['3222401L13Rik', 'Slc5a5', 'AC149090.1', 'Ccnd1', 'Ccnd2']
)
subclass_by_gene.head()

	3222401L13Rik	Slc5a5	AC149090.1	Ccnd1	Ccnd2
001 CLA-EPd-CTX Car3 Glut	NaN	NaN	1.614551	NaN	NaN
002 IT EP-CLA Glut	NaN	1.096541	NaN	-2.573835	-1.277657
003 L5/6 IT TPE-ENT Glut	NaN	1.253065	1.909579	-1.420208	NaN
004 L6 IT CTX Glut	NaN	NaN	1.535055	-1.5732	-1.371331
005 L5 IT CTX Glut	NaN	1.886562	1.530523	NaN	NaN

res = plot_heatmap(subclass_by_gene, 8, 25, cmap=plt.cm.RdBu_r, vmin=-4, vmax=4)
plt.show()

We can also plot the ageDE expression for the non-neuronal supertypes from Jin et al. Note that we show slightly different genes for the non-neuronal celltypes compared to our previous plot.

supertype_by_gene = get_genes_for_celltype_level(
    degene_df=degenes,
    level='supertype',
    degene_list=['3222401L13Rik', 'AC149090.1', 'Gm10076', 'Uba52', 'Ccnd2']
)
supertype_by_gene.head()

	3222401L13Rik	AC149090.1	Gm10076	Uba52	Ccnd2
1159 Astro-NT NN_1	NaN	1.606709	NaN	NaN	NaN
1160 Astro-NT NN_2	NaN	1.917127	NaN	NaN	NaN
1161 Astro-TE NN_1	NaN	2.299556	NaN	1.570618	-1.361851
1163 Astro-TE NN_3	1.117364	2.569296	NaN	1.169082	NaN
1165 Astro-TE NN_5	NaN	1.228538	1.000455	1.787092	NaN

res = plot_heatmap(supertype_by_gene, 8, 8, cmap=plt.cm.RdBu_r, vmin=-4, vmax=4)
plt.show()