-
Notifications
You must be signed in to change notification settings - Fork 9
/
Copy pathvariance_adjusted_permutations.py
159 lines (126 loc) · 6.43 KB
/
variance_adjusted_permutations.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
import argparse
import numpy as np
import pandas as pd
from scipy import stats
from tqdm import tqdm
from statsmodels.stats.multitest import multipletests
def get_args():
parser = argparse.ArgumentParser(description="calculate p values based on Romano method")
parser.add_argument("--dataname",help="name of dataset to use")
parser.add_argument("--suffix",help="suffix to save with")
parser.add_argument("--num_perms",type=int,help="number of permutations to run for")
parser.add_argument("--temp",action="store_true",help="use temp rijk file")
args = parser.parse_args()
return args
def calc_pval(var_df):
# calculate the inner sum that's subtracted
num = 0
denom = 0
for index, row in var_df.iterrows():
num += row["num_cells_ont"]*row["ont_median"]/row["ont_var"]
denom += row["num_cells_ont"]/row["ont_var"]
const = num/denom
# calculate the outer sum
sum_vals = 0
for index, row in var_df.iterrows():
sum_vals += (row["num_cells_ont"]/row["ont_var"])*(row["ont_median"] - const)**2
# return the chi^2 p value and the chi^2 statistic
return 1 - stats.chi2.cdf(sum_vals , var_df.shape[0] - 1), sum_vals
def get_var_df(sub_df, z_col, adj_var):
sub_df["num_cells_ont"] = sub_df["ontology"].map(sub_df.groupby("ontology")["cell"].nunique())
sub_df["ont_median"] = sub_df["ontology"].map(sub_df.groupby("ontology")[z_col].median())
sub_df["ont_var"] = sub_df["ontology"].map(sub_df.groupby("ontology")[z_col].var())
var_df = sub_df.drop_duplicates("ontology")[["ontology","ont_median","num_cells_ont","ont_var"]]
# don't need to remove cell types with variance 0 when we're adjusting variance
if not adj_var:
# remove ontologies with zero variance
var_df = var_df[var_df["ont_var"] > 0]
return var_df
def main():
alpha = 0.05
outpath = "scripts/output/variance_adjusted_permutations/"
args = get_args()
np.random.seed(123)
df = pd.read_parquet("scripts/output/rijk_zscore/{}_sym_SVD_normdonor{}.pq".format(args.dataname,args.suffix),columns=["geneR1A_uniq","ontology","cell","scZ","svd_z0","svd_z1","svd_z2","cell_gene","f0","f1","f2"])
df = df.drop_duplicates("cell_gene")
# subset to ontologies with > 20 cells
df["ontology_gene"] = df["ontology"] + df["geneR1A_uniq"]
df["num_ont_gene"] = df["ontology_gene"].map(df.groupby("ontology_gene")["cell_gene"].nunique())
df = df[df["num_ont_gene"] > 10]
z_cols = ["scZ","svd_z0","svd_z1","svd_z2"]
out = {"pval" : [], "geneR1A_uniq" : [], "num_onts" : [],"z_col" : [],"max_abs_median" : [], "Tn1" : []}
var_adj = 0.1
adj_var = True
perm_pval = True
if perm_pval:
out["perm_pval"] = []
for gene, sub_df in tqdm(df.groupby("geneR1A_uniq")):
for z_col in z_cols:
var_df = get_var_df(sub_df, z_col, adj_var)
if var_df.shape[0] > 1:
if adj_var:
var_df["ont_var"] = var_df["ont_var"] + var_adj
pval, Tn1 = calc_pval(var_df)
out["pval"].append(pval)
out["Tn1"].append(Tn1)
out["geneR1A_uniq"].append(gene)
out["num_onts"].append(var_df.shape[0])
out["z_col"].append(z_col)
out["max_abs_median"].append((var_df["ont_median"].abs()).max())
if perm_pval:
sub_df_perm = sub_df.copy()
if (pval < alpha):
Tn1_dist = []
# for i in range(args.num_perms):
while len(Tn1_dist) < args.num_perms:
sub_df_perm["ontology"] = np.random.permutation(sub_df_perm["ontology"])
var_df = get_var_df(sub_df_perm, z_col, adj_var)
if var_df.shape[0] > 1:
if adj_var:
var_df["ont_var"] = var_df["ont_var"] + var_adj
pval, Tn1_perm = calc_pval(var_df)
Tn1_dist.append(Tn1_perm)
out["perm_pval"].append(len([x for x in Tn1_dist if x < Tn1])/args.num_perms)
else:
out["perm_pval"].append(np.nan)
out_df = pd.DataFrame.from_dict(out)
out_df["perm_pval_inv"] = 1 - out_df["perm_pval"]
out_df["perm_pval2"] = 2*out_df[["perm_pval","perm_pval_inv"]].min(axis=1)
# adjust p values separately per z score
for z_col in z_cols:
out_df.loc[out_df["z_col"] == z_col, "pval_adj"] = multipletests(out_df.loc[out_df["z_col"] == z_col, "pval"],alpha, method="fdr_bh")[1]
out_df.loc[(out_df["z_col"] == z_col) & (~out_df["perm_pval2"].isna()), "perm_pval2_adj"] = multipletests(out_df.loc[(out_df["z_col"] == z_col) & (~out_df["perm_pval2"].isna()), "perm_pval2"],alpha, method="fdr_bh")[1]
out_df.to_csv("{}{}_outdf_{}{}.tsv".format(outpath,args.dataname, args.num_perms,args.suffix),sep="\t",index=False)
# out_df["pval_adj"] = multipletests(out_df["pval"],alpha, method="fdr_bh")[1]
# out_df["pval_adj"] = multipletests(out_df["pval"],alpha, method="fdr_bh")[1]
# out_df.loc[~out_df["perm_pval2"].isna(),"perm_pval2_adj"] = multipletests(out_df.loc[~out_df["perm_pval2"].isna(),"perm_pval2"], alpha, method = "fdr_bh")[1]
# reformat output
new_out = {"geneR1A_uniq" : [], "num_onts" : []}
for z_col in z_cols:
new_out["chi2_pval_adj_" + z_col] = []
new_out["perm_pval_adj_" + z_col] = []
new_out["max_abs_median_" + z_col] = []
new_out["perm_cdf_" + z_col] = []
for gene, gene_df in out_df.groupby("geneR1A_uniq"):
new_out["geneR1A_uniq"].append(gene)
new_out["num_onts"].append(gene_df["num_onts"].iloc[0])
temp_z_cols = []
for z_col, z_df in gene_df.groupby("z_col"):
new_out["chi2_pval_adj_" + z_col].append(z_df["pval_adj"].iloc[0])
new_out["perm_pval_adj_" + z_col].append(z_df["perm_pval2_adj"].iloc[0])
new_out["max_abs_median_" + z_col].append(z_df["max_abs_median"].iloc[0])
new_out["perm_cdf_" + z_col].append(z_df["perm_pval"].iloc[0])
temp_z_cols.append(z_col)
for z_col in [x for x in z_cols if x not in temp_z_cols]:
new_out["chi2_pval_adj_" + z_col].append(np.nan)
new_out["perm_pval_adj_" + z_col].append(np.nan)
new_out["max_abs_median_" + z_col].append(np.nan)
new_out["perm_cdf_" + z_col].append(np.nan)
new_out_df = pd.DataFrame.from_dict(new_out).sort_values("perm_pval_adj_scZ")
# add frac from SVD for each gene
df = df.drop_duplicates("geneR1A_uniq")
for i in range(3):
frac_dict = pd.Series(df["f" + str(i)].values,index=df.geneR1A_uniq).to_dict()
new_out_df["f" + str(i)] = new_out_df["geneR1A_uniq"].map(frac_dict)
new_out_df.to_csv("{}{}_pvals_{}{}.tsv".format(outpath,args.dataname, args.num_perms,args.suffix),sep="\t",index=False)
main()