Alteration of intercellular signaling in ulcerative colitis

Analysising single cell RNA-seq data for inter and intracellular interactions. The first step is to import the singel cell RNA-seq data file. The import file is the cell cluster averaged gene expression files. OmniPath is accesed by the python client.

import pandas as pd
import igraph as ig
import os
from itertools import permutations
import numpy as np

import omnipath as op

C:\Users\modos\.conda\envs\Single_cell_RNA_seq\lib\site-packages\requests\__init__.py:91: RequestsDependencyWarning: urllib3 (1.26.2) or chardet (3.0.4) doesn't match a supported version!
  RequestsDependencyWarning)

First lets check the OmniPAth version and data.

print(op.__server_version__)
print(op.options)

0.11.34
Options(url='https://omnipathdb.org', license=<License.ACADEMIC>, cache=<FileCache[size=3, path='C:\\Users\\modos\\.cache\\omnipathdb']>, autoload=True, convert_dtypes=True, num_retries=3, timeout=5.0, chunk_size=8196)

Now we can donwload OmniPath intercellular network and see what kind of columns are in the intercellular dataframe.

intercell_network = op.interactions.import_intercell_network()

intercell_network.head()

	index	source	target	is_stimulation	is_inhibition	consensus_direction	consensus_stimulation	consensus_inhibition	dip_url	curation_effort	...	category_source_intercell_target	uniprot_intercell_target	genesymbol_intercell_target	entity_type_intercell_target	consensus_score_intercell_target	transmitter_intercell_target	receiver_intercell_target	secreted_intercell_target	plasma_membrane_transmembrane_intercell_target	plasma_membrane_peripheral_intercell_target
0	0	P14416	P48995	True	False	True	True	False	nan	1	...	resource_specific	P48995	TRPC1	protein	1	False	True	False	False	False
1	2	P14416	P48995	True	False	True	True	False	nan	1	...	resource_specific	P48995	TRPC1	protein	3	False	True	False	False	False
2	5	P14416	P48995	True	False	True	True	False	nan	1	...	composite	P48995	TRPC1	protein	3	False	True	False	False	False
3	7	Q13255	P48995	True	False	True	True	False	nan	1	...	resource_specific	P48995	TRPC1	protein	1	False	True	False	False	False
4	9	Q13255	P48995	True	False	True	True	False	nan	1	...	resource_specific	P48995	TRPC1	protein	3	False	True	False	False	False

5 rows × 46 columns

intercell_network.columns

Index(['index', 'source', 'target', 'is_stimulation', 'is_inhibition',
       'consensus_direction', 'consensus_stimulation', 'consensus_inhibition',
       'dip_url', 'curation_effort', 'references', 'sources',
       'references_stripped', 'n_references', 'n_sources', 'n_primary_sources',
       'category_intercell_source', 'parent_intercell_source',
       'database_intercell_source', 'scope_intercell_source',
       'aspect_intercell_source', 'category_source_intercell_source',
       'uniprot_intercell_source', 'genesymbol_intercell_source',
       'entity_type_intercell_source', 'consensus_score_intercell_source',
       'transmitter_intercell_source', 'receiver_intercell_source',
       'secreted_intercell_source',
       'plasma_membrane_transmembrane_intercell_source',
       'plasma_membrane_peripheral_intercell_source',
       'category_intercell_target', 'parent_intercell_target',
       'database_intercell_target', 'scope_intercell_target',
       'aspect_intercell_target', 'category_source_intercell_target',
       'uniprot_intercell_target', 'genesymbol_intercell_target',
       'entity_type_intercell_target', 'consensus_score_intercell_target',
       'transmitter_intercell_target', 'receiver_intercell_target',
       'secreted_intercell_target',
       'plasma_membrane_transmembrane_intercell_target',
       'plasma_membrane_peripheral_intercell_target'],
      dtype='object')

intercell_network.shape

(30265, 46)

Here we can check what are the categories what we can use for intercellular interactions. We can use the intercellular interctions directly, hovewer it contain porteins which are invovled in the adhesion process and binds to the intracelluar domains of receptor proteins. That is why we suggest the fitlering as in the paper.

print (set(intercell_network["category_intercell_source"]))

{'desmosome', 'cell_surface_ligand', 'cell_surface_enzyme', 'secreted_receptor', 'ecm', 'receptor_regulator', 'adhesion', 'tight_junction', 'cell_adhesion', 'matrix_adhesion_regulator', 'cell_surface_peptidase', 'ecm_regulator', 'secreted_enzyme', 'ligand_regulator', 'gap_junction', 'ligand'}

print (set(intercell_network["category_intercell_target"]))

{'adherens_junction', 'receptor', 'ion_channel_regulator', 'desmosome', 'adhesion', 'tight_junction', 'cell_adhesion', 'matrix_adhesion', 'transporter', 'ion_channel', 'gap_junction'}

filtered_source_types = ["cell_surface_enzyme","cell_surface_ligand","ligand","secreted_enzyme","secreted_receptor",
           "adhesion","cell_adhesion","tight_junction","cell_surface_peptidase","gap_junction","desmosome"]
filtered_target_types = ["adhesion","receptor","cell_adhesion","tight_junction","gap_junction","ion_channel","transporter",
                "adherens_junction"]

filtered_intercell_network = intercell_network[intercell_network.category_intercell_source.isin(filtered_source_types)] 
filtered_intercell_network = filtered_intercell_network[filtered_intercell_network.
                                                        category_intercell_target.isin(filtered_target_types)]
filtered_intercell_network.shape

(20784, 46)

The next step is to preprocess the data files. The imput file is the cell type specfic expression. The data are from Smillie et al 2019 (https://pubmed.ncbi.nlm.nih.gov/31348891/). Each column is the average per cell type in transcript per million. From that we will build a simple cell-cell interaction count. It is based on expression tresholding. Our assumptation was that the interactions are important if they are exisit in any way, so we have chosen a realtievely low expression treshold. We invite the user for testing various tresholds in their own work. You will need for this the expression data which are in this example's folder.

df_cell_imm = pd.read_csv("imm_all_expression_condition.tsv", sep="\t", index_col=0, header=0)
df_cell_fib = pd.read_csv("fib_all_expression_condition.tsv",sep="\t", index_col=0, header=0)
df_cell_epi =  pd.read_csv("epi_all_expression_condition.tsv",sep="\t", index_col=0, header=0)
df_cell_exp = df_cell_imm.join(df_cell_fib, how="outer") #Keeping all genes
df_cell_exp = df_cell_exp.join(df_cell_epi, how="outer")

df_cell_exp.head()

	RNA.CD8..IELs_Healthy	RNA.CD8..LP_Healthy	RNA.CD4..Memory_Healthy	RNA.MT.hi_Healthy	RNA.Cycling.T_Healthy	RNA.NKs_Healthy	RNA.CD4..Activated.Fos.lo_Healthy	RNA.CD4..Activated.Fos.hi_Healthy	RNA.CD8..IELs_Uninflamed	RNA.MT.hi_Uninflamed	...	RNA.Immature.Goblet_Inflamed	RNA.Stem_Uninflamed	RNA.Immature.Enterocytes.2_Inflamed	RNA.Goblet_Inflamed	RNA.Tuft_Inflamed	RNA.Enterocytes_Inflamed	RNA.Best4..Enterocytes_Inflamed	RNA.Enteroendocrine_Inflamed	RNA.M.cells_Inflamed	RNA.M.cells_Uninflamed
7SK	0.030182	0.042158	0.016792	0.000000	0.000000	0.228991	0.000000	0.015092	0.030573	0.000000	...	0.00000	0.001679	0.000000	0.000000	7.005192	0.002465	0.007913	0.000000	0.000000	0.000000
A1BG	0.062424	0.025848	0.049498	0.208431	0.321939	0.000000	0.045158	0.018531	0.053037	0.000000	...	0.00000	0.013224	0.002473	0.089265	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
A1BG-AS1	0.276995	0.198592	0.138406	0.295042	0.000000	0.498747	0.130186	0.256156	0.305482	0.000000	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
A1CF	0.002978	0.005669	0.000000	0.000000	0.025840	0.000000	0.000000	0.000000	0.079150	0.000000	...	0.66123	0.211192	0.848589	0.187466	0.000000	0.486224	0.778587	2.653654	0.457293	0.737544
A2M	0.023545	0.214442	0.202609	0.167921	0.097767	0.545677	0.144894	0.166956	0.010304	0.243075	...	0.00000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000

5 rows × 153 columns

We will choose the mean -2SD of the whole data set. First we do a log2 based transformation.

df_cell_exp[df_cell_exp.values == 0] = "NaN"

df_cell_exp.head()

	RNA.CD8..IELs_Healthy	RNA.CD8..LP_Healthy	RNA.CD4..Memory_Healthy	RNA.MT.hi_Healthy	RNA.Cycling.T_Healthy	RNA.NKs_Healthy	RNA.CD4..Activated.Fos.lo_Healthy	RNA.CD4..Activated.Fos.hi_Healthy	RNA.CD8..IELs_Uninflamed	RNA.MT.hi_Uninflamed	...	RNA.Immature.Goblet_Inflamed	RNA.Stem_Uninflamed	RNA.Immature.Enterocytes.2_Inflamed	RNA.Goblet_Inflamed	RNA.Tuft_Inflamed	RNA.Enterocytes_Inflamed	RNA.Best4..Enterocytes_Inflamed	RNA.Enteroendocrine_Inflamed	RNA.M.cells_Inflamed	RNA.M.cells_Uninflamed
7SK	0.0301819	0.0421578	0.0167924	NaN	NaN	0.228991	NaN	0.0150918	0.0305725	NaN	...	NaN	0.00167886	NaN	NaN	7.00519	0.00246488	0.00791332	NaN	NaN	NaN
A1BG	0.062424	0.0258481	0.0494975	0.208431	0.321939	NaN	0.0451578	0.0185314	0.0530365	NaN	...	NaN	0.0132239	0.0024732	0.089265	NaN	NaN	NaN	NaN	NaN	NaN
A1BG-AS1	0.276995	0.198592	0.138406	0.295042	NaN	0.498747	0.130186	0.256156	0.305482	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
A1CF	0.00297762	0.00566872	NaN	NaN	0.0258403	NaN	NaN	NaN	0.0791505	NaN	...	0.66123	0.211192	0.848589	0.187466	NaN	0.486224	0.778587	2.65365	0.457293	0.737544
A2M	0.0235452	0.214442	0.202609	0.167921	0.0977667	0.545677	0.144894	0.166956	0.0103037	0.243075	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

5 rows × 153 columns

data_matrix = df_cell_exp.to_numpy()
data_matrix_log2 = np.log2(data_matrix.astype(float)) #We need to change the data file to floats
mean_cell = np.nanmean(data_matrix_log2)
std = np.nanstd(data_matrix_log2)

mean_cell, std

(-2.160708171397226, 3.00937762482671)

import seaborn as sns

We can check the distribuiton of the cell types. Here only for one.

data_matrix_log2

array([[-5.0501723 , -4.56805614, -5.89604351, ...,         nan,
                nan,         nan],
       [-4.00175632, -5.27379813, -4.33649916, ...,         nan,
                nan,         nan],
       [-1.85206773, -2.33211901, -2.8530207 , ...,         nan,
                nan,         nan],
       ...,
       [-3.47290816, -4.19215007, -4.40580567, ..., -2.06536459,
         0.35303725,  0.44592408],
       [-4.38586399, -2.27399124, -3.85313444, ..., -4.58601467,
        -8.31527547,         nan],
       [-2.5735855 , -1.76640068, -1.1066736 , ..., -4.14328163,
        -7.335609  ,         nan]])

sns.displot(data_matrix_log2[:,0])

<seaborn.axisgrid.FacetGrid at 0x19ee52d6c88>

After checking the histogram we can say that we will use the mean minus 2 standard deviation of the expressed genes.

data_matrix_log2 > (mean_cell-2*std)

C:\Users\modos\.conda\envs\Single_cell_RNA_seq\lib\site-packages\ipykernel_launcher.py:1: RuntimeWarning: invalid value encountered in greater
  """Entry point for launching an IPython kernel.

array([[ True,  True,  True, ..., False, False, False],
       [ True,  True,  True, ..., False, False, False],
       [ True,  True,  True, ..., False, False, False],
       ...,
       [ True,  True,  True, ...,  True,  True,  True],
       [ True,  True,  True, ...,  True, False, False],
       [ True,  True,  True, ...,  True,  True, False]])

For simplicity we call "NaN" each number which is below the treshold.

df_cell_log2 = pd.DataFrame(data=data_matrix_log2, index=df_cell_exp.index ,columns=df_cell_exp.columns)

df_cell_log2.head()

	RNA.CD8..IELs_Healthy	RNA.CD8..LP_Healthy	RNA.CD4..Memory_Healthy	RNA.MT.hi_Healthy	RNA.Cycling.T_Healthy	RNA.NKs_Healthy	RNA.CD4..Activated.Fos.lo_Healthy	RNA.CD4..Activated.Fos.hi_Healthy	RNA.CD8..IELs_Uninflamed	RNA.MT.hi_Uninflamed	...	RNA.Immature.Goblet_Inflamed	RNA.Stem_Uninflamed	RNA.Immature.Enterocytes.2_Inflamed	RNA.Goblet_Inflamed	RNA.Tuft_Inflamed	RNA.Enterocytes_Inflamed	RNA.Best4..Enterocytes_Inflamed	RNA.Enteroendocrine_Inflamed	RNA.M.cells_Inflamed	RNA.M.cells_Uninflamed
7SK	-5.050172	-4.568056	-5.896044	NaN	NaN	-2.126639	NaN	-6.050093	-5.031619	NaN	...	NaN	-9.218304	NaN	NaN	2.808425	-8.664267	-6.981502	NaN	NaN	NaN
A1BG	-4.001756	-5.273798	-4.336499	-2.262358	-1.635142	NaN	-4.468881	-5.753884	-4.236870	NaN	...	NaN	-6.240706	-8.659406	-3.485761	NaN	NaN	NaN	NaN	NaN	NaN
A1BG-AS1	-1.852068	-2.332119	-2.853021	-1.761008	NaN	-1.003621	-2.941349	-1.964903	-1.710843	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
A1CF	-8.391626	-7.462761	NaN	NaN	-5.274231	NaN	NaN	NaN	-3.659258	NaN	...	-0.596775	-2.243375	-0.236862	-2.415296	NaN	-1.040307	-0.361069	1.40798	-1.12881	-0.4392
A2M	-5.408426	-2.221338	-2.303227	-2.574148	-3.354513	-0.873881	-2.786930	-2.582460	-6.600697	-2.040528	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

5 rows × 153 columns

We will call every genee which is expressed below the treshold with "NaN".

df_cell_log2[df_cell_log2.values < (mean_cell-2*std)] = "NaN"

C:\Users\modos\.conda\envs\Single_cell_RNA_seq\lib\site-packages\ipykernel_launcher.py:1: RuntimeWarning: invalid value encountered in less
  """Entry point for launching an IPython kernel.

Let's see what we hjave done and the avaialbe cell types.

df_cell_log2.head()

	RNA.CD8..IELs_Healthy	RNA.CD8..LP_Healthy	RNA.CD4..Memory_Healthy	RNA.MT.hi_Healthy	RNA.Cycling.T_Healthy	RNA.NKs_Healthy	RNA.CD4..Activated.Fos.lo_Healthy	RNA.CD4..Activated.Fos.hi_Healthy	RNA.CD8..IELs_Uninflamed	RNA.MT.hi_Uninflamed	...	RNA.Immature.Goblet_Inflamed	RNA.Stem_Uninflamed	RNA.Immature.Enterocytes.2_Inflamed	RNA.Goblet_Inflamed	RNA.Tuft_Inflamed	RNA.Enterocytes_Inflamed	RNA.Best4..Enterocytes_Inflamed	RNA.Enteroendocrine_Inflamed	RNA.M.cells_Inflamed	RNA.M.cells_Uninflamed
7SK	-5.05017	-4.56806	-5.896044	NaN	NaN	-2.126639	NaN	-6.05009	-5.031619	NaN	...	NaN	NaN	NaN	NaN	2.808425	NaN	-6.9815	NaN	NaN	NaN
A1BG	-4.00176	-5.2738	-4.336499	-2.262358	-1.635142	NaN	-4.46888	-5.75388	-4.236870	NaN	...	NaN	-6.24071	NaN	-3.48576	NaN	NaN	NaN	NaN	NaN	NaN
A1BG-AS1	-1.85207	-2.33212	-2.853021	-1.761008	NaN	-1.003621	-2.94135	-1.9649	-1.710843	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
A1CF	NaN	-7.46276	NaN	NaN	-5.274231	NaN	NaN	NaN	-3.659258	NaN	...	-0.596775	-2.24338	-0.236862	-2.4153	NaN	-1.04031	-0.361069	1.40798	-1.12881	-0.4392
A2M	-5.40843	-2.22134	-2.303227	-2.574148	-3.354513	-0.873881	-2.78693	-2.58246	-6.600697	-2.040528	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

5 rows × 153 columns

print(set(df_cell_log2.columns))

{'RNA.Macrophages_Uninflamed', 'RNA.CD4..PD1._Inflamed', 'RNA.ILCs_Inflamed', 'RNA.Cycling.TA_Healthy', 'RNA.TA.1_Inflamed', 'RNA.WNT2B..Fos.lo.1_Uninflamed', 'RNA.Tregs_Healthy', 'RNA.Best4..Enterocytes_Inflamed', 'RNA.Enterocyte.Progenitors_Inflamed', 'RNA.Microvascular_Inflamed', 'RNA.Goblet_Inflamed', 'RNA.CD4..Activated.Fos.hi_Healthy', 'RNA.CD8..IELs_Inflamed', 'RNA.WNT5B..2_Uninflamed', 'RNA.Inflammatory.Fibroblasts_Inflamed', 'RNA.CD8..IELs_Healthy', 'RNA.DC2_Inflamed', 'RNA.Tregs_Uninflamed', 'RNA.Immature.Enterocytes.1_Inflamed', 'RNA.WNT5B..2_Healthy', 'RNA.Follicular_Uninflamed', 'RNA.Inflammatory.Monocytes_Uninflamed', 'RNA.Microvascular_Uninflamed', 'RNA.Plasma_Uninflamed', 'RNA.CD8..IL17._Inflamed', 'RNA.CD8..IELs_Uninflamed', 'RNA.DC1_Inflamed', 'RNA.M.cells_Uninflamed', 'RNA.WNT5B..1_Inflamed', 'RNA.WNT2B..Fos.lo.2_Uninflamed', 'RNA.CD4..Activated.Fos.lo_Uninflamed', 'RNA.Cycling.Monocytes_Inflamed', 'RNA.NKs_Healthy', 'RNA.Secretory.TA_Inflamed', 'RNA.GC_Healthy', 'RNA.Cycling.B_Inflamed', 'RNA.Inflammatory.Monocytes_Healthy', 'RNA.MT.hi_Inflamed', 'RNA.Post.capillary.Venules_Inflamed', 'RNA.Cycling.T_Healthy', 'RNA.WNT2B..Fos.hi_Inflamed', 'RNA.RSPO3._Inflamed', 'RNA.Cycling.TA_Uninflamed', 'RNA.Enterocyte.Progenitors_Healthy', 'RNA.CD69..Mast_Inflamed', 'RNA.TA.2_Inflamed', 'RNA.TA.2_Healthy', 'RNA.DC1_Uninflamed', 'RNA.Endothelial_Inflamed', 'RNA.WNT5B..2_Inflamed', 'RNA.Pericytes_Healthy', 'RNA.Inflammatory.Monocytes_Inflamed', 'RNA.Inflammatory.Fibroblasts_Uninflamed', 'RNA.Microvascular_Healthy', 'RNA.M.cells_Inflamed', 'RNA.Glia_Healthy', 'RNA.MT.hi_Uninflamed', 'RNA.DC2_Healthy', 'RNA.Macrophages_Healthy', 'RNA.Macrophages_Inflamed', 'RNA.CD4..PD1._Healthy', 'RNA.Cycling.T_Uninflamed', 'RNA.CD69..Mast_Healthy', 'RNA.Enteroendocrine_Inflamed', 'RNA.Tregs_Inflamed', 'RNA.Secretory.TA_Healthy', 'RNA.CD4..Activated.Fos.hi_Uninflamed', 'RNA.Follicular_Inflamed', 'RNA.Plasma_Healthy', 'RNA.Immature.Enterocytes.1_Uninflamed', 'RNA.Glia_Inflamed', 'RNA.Best4..Enterocytes_Healthy', 'RNA.Best4..Enterocytes_Uninflamed', 'RNA.Stem_Inflamed', 'RNA.Tuft_Inflamed', 'RNA.Enterocytes_Uninflamed', 'RNA.Goblet_Uninflamed', 'RNA.CD4..Memory_Healthy', 'RNA.RSPO3._Uninflamed', 'RNA.Immature.Goblet_Inflamed', 'RNA.Enterocytes_Healthy', 'RNA.CD4..Activated.Fos.hi_Inflamed', 'RNA.MT.hi_Healthy', 'RNA.CD8..IL17._Healthy', 'RNA.WNT2B..Fos.lo.2_Inflamed', 'RNA.Goblet_Healthy', 'RNA.Cycling.B_Uninflamed', 'RNA.Cycling.Monocytes_Healthy', 'RNA.Pericytes_Inflamed', 'RNA.Immature.Enterocytes.2_Uninflamed', 'RNA.ILCs_Uninflamed', 'RNA.WNT2B..Fos.lo.1_Inflamed', 'RNA.Myofibroblasts_Healthy', 'RNA.CD4..Activated.Fos.lo_Healthy', 'RNA.Immature.Enterocytes.2_Healthy', 'RNA.WNT2B..Fos.lo.2_Healthy', 'RNA.M.cells_Healthy', 'RNA.CD8..IL17._Uninflamed', 'RNA.WNT2B..Fos.hi_Uninflamed', 'RNA.DC2_Uninflamed', 'RNA.TA.1_Uninflamed', 'RNA.Cycling.TA_Inflamed', 'RNA.Immature.Goblet_Healthy', 'RNA.Secretory.TA_Uninflamed', 'RNA.CD4..Memory_Uninflamed', 'RNA.Enteroendocrine_Uninflamed', 'RNA.Tuft_Uninflamed', 'RNA.Plasma_Inflamed', 'RNA.CD8..LP_Healthy', 'RNA.Stem_Uninflamed', 'RNA.Glia_Uninflamed', 'RNA.Follicular_Healthy', 'RNA.NKs_Uninflamed', 'RNA.DC1_Healthy', 'RNA.WNT5B..1_Uninflamed', 'RNA.CD4..Activated.Fos.lo_Inflamed', 'RNA.CD69..Mast_Healthy.1', 'RNA.NKs_Inflamed', 'RNA.Stem_Healthy', 'RNA.Inflammatory.Fibroblasts_Healthy', 'RNA.CD8..LP_Inflamed', 'RNA.CD69..Mast_Uninflamed.1', 'RNA.GC_Inflamed', 'RNA.Enterocyte.Progenitors_Uninflamed', 'RNA.GC_Uninflamed', 'RNA.Enterocytes_Inflamed', 'RNA.CD69..Mast_Inflamed.1', 'RNA.WNT2B..Fos.lo.1_Healthy', 'RNA.ILCs_Healthy', 'RNA.Post.capillary.Venules_Healthy', 'RNA.CD4..Memory_Inflamed', 'RNA.Myofibroblasts_Inflamed', 'RNA.WNT2B..Fos.hi_Healthy', 'RNA.WNT5B..1_Healthy', 'RNA.RSPO3._Healthy', 'RNA.Cycling.B_Healthy', 'RNA.CD8..LP_Uninflamed', 'RNA.Enteroendocrine_Healthy', 'RNA.TA.1_Healthy', 'RNA.Tuft_Healthy', 'RNA.Post.capillary.Venules_Uninflamed', 'RNA.Myofibroblasts_Uninflamed', 'RNA.Immature.Enterocytes.2_Inflamed', 'RNA.TA.2_Uninflamed', 'RNA.Immature.Goblet_Uninflamed', 'RNA.Endothelial_Uninflamed', 'RNA.Cycling.Monocytes_Uninflamed', 'RNA.Pericytes_Uninflamed', 'RNA.Immature.Enterocytes.1_Healthy', 'RNA.Cycling.T_Inflamed', 'RNA.CD69..Mast_Uninflamed', 'RNA.CD4..PD1._Uninflamed', 'RNA.Endothelial_Healthy'}

Next we select healthy/uninflamed average expression data in the 5 selected cell types.

df_macrophage_healthy = df_cell_log2['RNA.Macrophages_Healthy']
df_macrophage_uninflamed = df_cell_log2['RNA.Macrophages_Uninflamed']

df_DC1_healthy = df_cell_log2['RNA.DC1_Healthy']
df_DC1_uninflamed = df_cell_log2['RNA.DC1_Uninflamed']

df_Treg_healthy = df_cell_log2['RNA.Tregs_Healthy']
df_Treg_uninflamed = df_cell_log2['RNA.Tregs_Uninflamed'] 

df_myofibroblast_healthy = df_cell_log2['RNA.Myofibroblasts_Healthy']
df_myofibroblast_uninflamed = df_cell_log2['RNA.Myofibroblasts_Uninflamed']

df_goblet_healthy = df_cell_log2['RNA.Goblet_Healthy']
df_goblet_uninflamed = df_cell_log2['RNA.Goblet_Uninflamed']

For calcualting the intercellular interactions we can store theese dataframes in a dictionarry.

healthy_cell_types = {"DC1": df_DC1_healthy,"Macrophage": df_macrophage_healthy, "Goblet": df_goblet_healthy,
                      "Myofibroblast": df_myofibroblast_healthy, 'Treg': df_Treg_healthy}

uninflamed_cell_types = {"DC1": df_DC1_uninflamed,"Macrophage": df_macrophage_uninflamed, "Goblet": df_goblet_uninflamed,
                         "Myofibroblast": df_myofibroblast_uninflamed, 'Treg': df_Treg_uninflamed} 

Next we can use the intercellular interactions.

# create dictionaries for source-target interactions and their annotations from OmniPath
interactions = {}
interaction_annotation = {}

for i, interaction in filtered_intercell_network.iterrows():
    if interaction["genesymbol_intercell_source"] not in interactions:
        interactions[interaction["genesymbol_intercell_source"]] = []
    if interaction["genesymbol_intercell_target"] not in  interactions[interaction["genesymbol_intercell_source"]]:
        interactions[interaction["genesymbol_intercell_source"]].append(interaction["genesymbol_intercell_target"])
    source_tupple = (interaction["genesymbol_intercell_source"],interaction["category_intercell_source"])
    target_tupple = (interaction["genesymbol_intercell_target"],interaction["category_intercell_target"])
    if source_tupple not in interaction_annotation:
        interaction_annotation[source_tupple] = []
    if target_tupple not in interaction_annotation[source_tupple]:
        interaction_annotation[source_tupple].append(target_tupple)

Create all of the possible interactions of cells (independently from the condition) Important: directionality (A-B and B-A are different)

def create_interactions(cells, healthy_cell_types,interactions):
    healthy_interactions = set()
    for source in healthy_cell_types[cells[0]].keys():
            # selecting those ones which play role in intercellular communication as a transmitter
            if source in interactions.keys():
                # iterating through target cell type expressed genes
                for target in healthy_cell_types[cells[1]].keys():

                    # selecting those ones which play role in intercellular communication as a receiver
                    if target in interactions[source]:
                        if str(healthy_cell_types[cells[1]][target]) != 'nan' and str(healthy_cell_types[cells[0]][source]) != 'nan':
                            healthy_interactions.add((source, target))
    return healthy_interactions
    

#Creating dictinaries dataframe for healthy and UC interactions
healthy_intercell_tupplellist = []
uc_intercell_tupplelist = []

for i in healthy_cell_types.keys():
    for j in healthy_cell_types.keys():
        if i!=j:
            print(i,j)
            # storing cell-type specific connections in sets
            healthy_interactions = create_interactions([i,j], healthy_cell_types, interactions)
            UC_interactions = create_interactions([i,j], uninflamed_cell_types,interactions)
            # writing out the condition specific interactions
            healthy_only = healthy_interactions.difference(UC_interactions)
            UC_only = UC_interactions.difference(healthy_interactions)
            #Adding the intercellular interactions
            healthy_intercell_tupplellist.append((i,j,len(healthy_only)))
            uc_intercell_tupplelist.append((i,j,len(UC_only)))

            with open(i + "_" + j + "_healthy_only.txt", 'w') as output_file_1:
                #print(UC_only)
                for inter in healthy_only:
                    for source_annotation in interaction_annotation:
                        if inter[0] == source_annotation[0]:
                            for target_annotation in interaction_annotation[source_annotation]:
                                if inter[1] == target_annotation[0]:
                                    output_file_1.write(source_annotation[0] + "," + source_annotation[1] + "," 
                                                        + target_annotation[0] + "," + target_annotation[1] + "\n")
            with open(i + "_" + j + "_UC_only.txt", 'w') as output_file_2:
                for inter in UC_only:
                        for source_annotation in interaction_annotation:
                            if inter[0] == source_annotation[0]:
                                for target_annotation in interaction_annotation[source_annotation]:
                                    if inter[1] == target_annotation[0]:
                                        output_file_2.write(source_annotation[0] + "," + source_annotation[1] + ","
                                                            + target_annotation[0] + "," + target_annotation[1] + "\n")

DC1 Macrophage
DC1 Goblet
DC1 Myofibroblast
DC1 Treg
Macrophage DC1
Macrophage Goblet
Macrophage Myofibroblast
Macrophage Treg
Goblet DC1
Goblet Macrophage
Goblet Myofibroblast
Goblet Treg
Myofibroblast DC1
Myofibroblast Macrophage
Myofibroblast Goblet
Myofibroblast Treg
Treg DC1
Treg Macrophage
Treg Goblet
Treg Myofibroblast

Next we make the graphs of the interactions between the cells.

g_healthy = ig.Graph.TupleList(healthy_intercell_tupplellist, weights=True, directed=True)
g_uc = ig.Graph.TupleList(uc_intercell_tupplelist, weights=True, directed=True)

healthy_intercell_tupplellist

[('DC1', 'Macrophage', 912),
 ('DC1', 'Goblet', 781),
 ('DC1', 'Myofibroblast', 731),
 ('DC1', 'Treg', 557),
 ('Macrophage', 'DC1', 568),
 ('Macrophage', 'Goblet', 689),
 ('Macrophage', 'Myofibroblast', 617),
 ('Macrophage', 'Treg', 501),
 ('Goblet', 'DC1', 457),
 ('Goblet', 'Macrophage', 652),
 ('Goblet', 'Myofibroblast', 483),
 ('Goblet', 'Treg', 438),
 ('Myofibroblast', 'DC1', 437),
 ('Myofibroblast', 'Macrophage', 558),
 ('Myofibroblast', 'Goblet', 483),
 ('Myofibroblast', 'Treg', 387),
 ('Treg', 'DC1', 491),
 ('Treg', 'Macrophage', 800),
 ('Treg', 'Goblet', 682),
 ('Treg', 'Myofibroblast', 616)]

Now we can visualise the interactions. Igraph contain the visulasitation parameters as a dictionarry. You can also wirte out the edgelist to visualise it in cytoscape.

g_healthy.es

<igraph.EdgeSeq at 0x19ee53e1948>

visual_style = {}

Below are the iGraph visual parameters which you can play with for your intercellular network.

# Curve the edges
visual_style["edge_curved"] = True
# Set the layout
my_layout = g_healthy.layout_circle(order=["Goblet","Treg","Myofibroblast","Macrophage","DC1"])
visual_style["layout"] = my_layout
#Add annoation
visual_style["vertex_label"] = g_healthy.vs["name"]
#Calcualte the edge wheight relative the number of edges
visual_style["edge_width"] = 0.01 * np.array(g_healthy.es["weight"])

#Setting the vertex visualistation parameters -colours and label position
visual_style["vertex_label_size"] = 30
visual_style["vertex_color"] = "grey"
visual_style["vertex_label_dist"] = 0.75
visual_style["vertex_size"] = 40

#Setting the edge visualisation parameters
visual_style["edge_color"] = "#004C66" #hexa code for blue
visual_style["edge_arrow_size"] = 2

#Setting the visualistation place. igraph need margin yo work
visual_style["bbox"] = (600, 600)
visual_style["margin"] = 100

ig.plot(g_healthy, "healthy.png", **visual_style)

Now we can do the same for the uninflmaed UC cells.

# Curve the edges
visual_style["edge_curved"] = True
# Set the layout - we keep the same layout for comapriosn
visual_style["layout"] = my_layout
#Add annoation
visual_style["vertex_label"] = g_uc.vs["name"]
#Calcualte the edge wheight relative the number of edges
visual_style["edge_width"] = 0.01 * np.array(g_uc.es["weight"])

#Setting the vertex visualistation parameters -colours and label position
visual_style["vertex_label_size"] = 30
visual_style["vertex_color"] = "grey"
visual_style["vertex_label_dist"] = 0.75
visual_style["vertex_size"] = 40

#Setting the edge visualisation parameters
visual_style["edge_color"] = "#CE1612" #hexa code for red
visual_style["edge_arrow_size"] = 2

#Setting the visualistation place. igraph need margin yo work
visual_style["bbox"] = (600, 600)
visual_style["margin"] = 100

ig.plot(g_uc, "UC_uninflamed.png", **visual_style)

Below you can see what kind of data we have used.

import types
def imports():
    for name, val in globals().items():
        if isinstance(val, types.ModuleType):
            yield val.__name__
list(imports())

['builtins',
 'builtins',
 'pandas',
 'igraph',
 'os',
 'numpy',
 'omnipath',
 'seaborn',
 'types']

print('\n'.join(f'{m.__name__} {m.__version__}' for m in globals().values() if getattr(m, '__version__', None)))

pandas 1.1.4
igraph 0.8.3
numpy 1.18.5
omnipath 1.0.0
seaborn 0.11.0