zh1peng / allenbraintool

gene vs. t-maps

From reading this A practical guide to linking brain-wide gene expression and neuroimaging data paper,there are so many things need to be taken into account, so direct comparision between gene expression agains statastical maps may be not appraporiate! Use with big caution.

Allen Brain Atlas tool

Need a new documentation and webpage for this.

1. get gene expression from Allen Brain API (http://api.brain-map.org/api/v2/data/query.json)

2. average or pca gene expression using muliple

3. corrected mni for each well-id

4. get t-value from mni (average with r)

5. random gene selection

6. bootstrap on spearman correlation, which allow to build a null model to test if the target gene is 'real' correlate with the t-map.

7. find nii files

1-4 features are from alleninf (https://github.com/chrisfilo/alleninf), and changed to python 3x.

Before use

1.change line 173 to the gene list (i.e. gene_list_csv) available from allen brain atlas. This file is converted from official pdf documentation. Some genes are not available when using the toolbox to retreive.

2.change line 91 (csvfile) to the corrected_mni_coordinates.csv. This csv file is from https://github.com/chrisfilo/alleninf.

Example1 Get some random gene and run bootstrap

import pandas as pd
import os
from zhipeng_alleninf import *

all_file, file_names = find_nii(
    r"-----------------nii file dir-------------------")

gene_list = random_gene(3, 'DRD1')
all_gene_expression, well_ids = get_expression_values_from_gene_list(gene_list)
stat_map = all_file[0]
test, data_table = boostrap_nii_vs_gene_list(stat_map, well_ids, all_gene_expression, boot_n=500)
# test is [r1 r2 r3 r4..rn]*bootstrap_n results.
# rn is spearman r for nth gene against 1st column which is the nii values.
# use np.percentile to find percentile

Example2 Get gene of interest and get data table

import pandas as pd
import os
from zhipeng_alleninf import *

all_file, file_names = find_nii(
    r"--------------------nii file dir-------------------")
gene_list = ['DRD1', 'DRD2', 'DRD3']

all_gene_expression, well_ids = get_expression_values_from_gene_list(gene_list)

stat_map = all_file[0] #select the 1st stat map

data_table = boostrap_nii_vs_gene_list(stat_map, well_ids, all_gene_expression, boot_n=0)
#if boot_strap set to 0, only return data_table

# how to plot
import pylab as plt
import seaborn as sns
from scipy.stats.stats import pearsonr, spearmanr
labels = list(data_table.keys())
sns_plot1 = sns.jointplot(labels[0], labels[1], data_table.iloc[:, [0, 1]], kind="reg")
sns_plot2 = sns.jointplot(labels[0], labels[1], data_table.iloc[:,
                                                                [0, 1]], stat_func=spearmanr, kind="reg")

sns_plot.savefig(name+'results.tiff',dpi=600)

code snips on how to use

import pandas as pd
import os
from zhipeng_alleninf import *

all_file, file_names = find_nii(
    r'C:\Users\Zhipeng\Desktop\nii test')

gene_list = random_gene(4, 'CNR1')
all_gene_expression, well_ids1 = get_expression_values_from_gene_list(gene_list)

all_gene_expression=pd.concat([all_gene_expression,all_gene_expression1],axis=1)

stat_map = all_file[0]
test, data_table = boostrap_nii_vs_gene_list(stat_map, well_ids, all_gene_expression, boot_n=5000)

#all_r=np.array(test)
np.savetxt('random_gene_r_values.csv', all_r,delimiter=',')
data_table.to_csv('random_gene_expression_values.csv')
all_r=np.array(test).reshape(-1,1)
thres_95=np.percentile(all_r,95)
np.percentile(all_r,10)




gene_list=['CNR1']
all_gene_expression_CNR1, well_ids = get_expression_values_from_gene_list(gene_list)

stat_map = all_file[0] #select the 1st stat map

data_table_cor = boostrap_nii_vs_gene_list(stat_map, well_ids, all_gene_expression_CNR1, boot_n=0)
#if boot_strap set to 0, only return data_table

# how to plot
import pylab as plt
import seaborn as sns
from scipy.stats.stats import pearsonr, spearmanr
labels = list(data_table_cor.keys())
sns_plot1 = sns.jointplot(labels[0], labels[1], data_table_cor.iloc[:, [0, 1]], kind="reg")
markers='x'
sns_plot2 = sns.jointplot(labels[0], labels[1], data_table_cor.iloc[:,[0, 1]], 
                          stat_func=spearmanr, kind="reg",joint_kws={'marker':markers})
sns_plot2.savefig(r'C:\Users\Zhipeng\Desktop\nii test\AllenBrainTool-master\results.png',dpi=600)
data_table_cor.to_csv('CNR1_expression_values.csv')