Figure 3J - Supplementary Tables#

Import packages#

[1]:

import pandas as pd
import scdiffeq as sdq
import scipy.stats

Load results#

[2]:

Perturbed = sdq.io.read_pickle("./Spi1.perturb.z_scan.pkl")

Loop through the results, concat into table#

[3]:

list_of_dfs = []
for version, result in Perturbed.items():
    for z, expt in result.items():
        expt.stats['version'] = version
        expt.stats['z'] = z
        list_of_dfs.append(expt.stats)
df = pd.concat(list_of_dfs)
df

[3]:

	lfc	lfc_std	pval	version	z
Baso	0.012356	0.152108	0.847814	version_0	-10.0
Ccr7_DC	-3.786314	8.466456	0.373901	version_0	-10.0
Eos	-0.483007	0.360628	0.072644	version_0	-10.0
Lymphoid	3.986314	15.800427	0.607584	version_0	-10.0
Mast	0.031224	0.144437	0.621027	version_0	-10.0
...	...	...	...	...	...
Mast	-0.029032	0.223107	0.682293	version_4	10.0
Meg	-0.336100	0.644422	0.206413	version_4	10.0
Monocyte	0.170410	0.081715	0.002381	version_4	10.0
Neutrophil	0.066611	0.074323	0.050309	version_4	10.0
Undifferentiated	-0.088646	0.051462	0.018303	version_4	10.0

406 rows × 5 columns

Write to csv#

[4]:

df.to_csv("supplementary_table.accompanies_3J.csv")

Compute correlations#

Annotate the Figure 3J with the results

[5]:

grouped = df.reset_index().rename({"index": "fate"}, axis = 1).groupby("fate")
for fate in ['Neutrophil', 'Monocyte', 'Baso']:
    fate_group = grouped.get_group(fate)
    corr, pval = scipy.stats.pearsonr(fate_group['z'].values, fate_group['lfc'].values)
    print("{:<10} | corr: {:<6} [pval: {:.3e}]".format(fate, round(corr, 3), pval))

Neutrophil | corr: 0.684  [pval: 2.280e-07]
Monocyte   | corr: 0.883  [pval: 1.034e-15]
Baso       | corr: -0.906 [pval: 1.182e-17]