-
Notifications
You must be signed in to change notification settings - Fork 0
/
rf_tree_chunk.py
177 lines (143 loc) · 7.09 KB
/
rf_tree_chunk.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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
import numpy as np
from sklearn.svm import SVR
#from sklearn.model_selection import train_test_split
import pandas as pd
from sklearn.model_selection import cross_val_score
from statistics import mean
import math
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import scale
from pandas import DataFrame
import pickle
from sklearn.ensemble import RandomForestRegressor
from sklearn.svm import SVR
from sklearn.neighbors import KNeighborsRegressor
import time
from scipy import stats
import argparse
from sklearn.linear_model import ElasticNet
import gzip
parser = argparse.ArgumentParser()
parser.add_argument("chr", action="store", help="put chromosome no")
parser.add_argument("chunk", action="store", help="put chromosome chunk no")
args = parser.parse_args()
chrom = args.chr
chrom = str(chrom)
chunk = str(args.chunk)
pop = "AFA"
#important functions needed
def get_filtered_snp_annot (snpfilepath):
snpanot = pd.read_csv(snpfilepath, sep="\t")
#snpanot = snpanot[((snpanot["refAllele"]=="A") & (snpanot["effectAllele"]=="C")) | ((snpanot["refAllele"]=="C") & (snpanot["effectAllele"]=="A")) | ((snpanot["refAllele"]=="A") & (snpanot["effectAllele"]=="G")) | ((snpanot["refAllele"]=="G") & (snpanot["effectAllele"]=="A")) | ((snpanot["refAllele"]=="T") & (snpanot["effectAllele"]=="G")) | ((snpanot["refAllele"]=="G") & (snpanot["effectAllele"]=="T")) | ((snpanot["refAllele"]=="T") & (snpanot["effectAllele"]=="C")) | ((snpanot["refAllele"]=="C") & (snpanot["effectAllele"]=="T"))]
snpanot = snpanot[(((snpanot["refAllele"]=="A") & (snpanot["effectAllele"]=="C")) | ((snpanot["refAllele"]=="C") & (snpanot["effectAllele"]=="A")) | ((snpanot["refAllele"]=="A") & (snpanot["effectAllele"]=="G")) | ((snpanot["refAllele"]=="G") & (snpanot["effectAllele"]=="A")) | ((snpanot["refAllele"]=="T") & (snpanot["effectAllele"]=="G")) | ((snpanot["refAllele"]=="G") & (snpanot["effectAllele"]=="T")) | ((snpanot["refAllele"]=="T") & (snpanot["effectAllele"]=="C")) | ((snpanot["refAllele"]=="C") & (snpanot["effectAllele"]=="T"))) & (snpanot["rsid"].notna())]
snpanot = snpanot.drop_duplicates(["varID"])
return snpanot
def get_gene_annotation (gene_anot_filepath, chrom, gene_types=["protein_coding"]):
gene_anot = pd.read_csv(gene_anot_filepath, sep="\t")
gene_anot = gene_anot[(gene_anot["chr"]==str(chrom)) & (gene_anot["gene_type"].isin(gene_types))]
return gene_anot
def get_gene_type (gene_anot, gene):
gene_type = gene_anot[gene_anot["gene_id"]==gene]
gene_type = gene_type.iloc[0,5]
return gene_type
def get_gene_name (gene_anot, gene):
gene_name = gene_anot[gene_anot["gene_id"]==gene]
gene_name = gene_name.iloc[0,2]
return gene_name
def get_gene_coords (gene_anot, gene):
gene_type = gene_anot[gene_anot["gene_id"]==gene]
gene_coord = [gene_type.iloc[0,3], gene_type.iloc[0,4]]
return gene_coord
def get_covariates (cov_filepath):
cov = pd.read_csv(cov_filepath, sep=" ")
cov = cov.set_index("IID") #make IID to be the row names
cov.index.names = [None] # remove the iid name from the row
pc = ["PC1", "PC2", "PC3"] #a list of the PCs to retain
cov = cov[pc]
return cov
def get_gene_expression(gene_expression_file_name, gene_annot):
expr_df = pd.read_csv(gene_expression_file_name, header = 0, index_col = 0, delimiter='\t')
expr_df = expr_df.T
inter = list(set(gene_annot['gene_id']).intersection(set(expr_df.columns)))
#print(len(inter))
expr_df = expr_df.loc[:, inter ]
return expr_df
def adjust_for_covariates (expr_vec, cov_df):
reg = LinearRegression().fit(cov_df, expr_vec)
ypred = reg.predict(cov_df)
residuals = expr_vec - ypred
residuals = scale(residuals)
return residuals
def get_maf_filtered_genotype(genotype_file_name, maf):
gt_df = pd.read_csv(genotype_file_name, 'r', header = 0, index_col = 0,delimiter='\t')
effect_allele_freqs = gt_df.mean(axis=1)
effect_allele_freqs = [ x / 2 for x in effect_allele_freqs ]
effect_allele_boolean = pd.Series([ ((x >= maf) & (x <= (1 - maf))) for x in effect_allele_freqs ]).values
gt_df = gt_df.loc[ effect_allele_boolean ]
gt_df = gt_df.T
return gt_df
def get_cis_genotype (gt_df, snp_annot, coords, cis_window=1000000):
snp_info = snpannot[(snpannot['pos'] >= (coords[0] - cis_window)) & (snpannot['rsid'].notna()) & (snpannot['pos'] <= (coords[1] + cis_window))]
if len(snp_info) == 0:
return 0
else:
gtdf_col = list(gt_df.columns)
snpinfo_col = list(snp_info["varID"])
intersect = snps_intersect(gtdf_col, snpinfo_col) #this function was defined earlier
cis_gt = gt_df[intersect]
return cis_gt
def calc_R2 (y, y_pred):
tss = 0
rss = 0
for i in range(len(y)):
tss = tss + (y[i])**2
rss = rss + (((y[i]) - (y_pred[i]))**2)
tss = float(tss)
rss = float(rss)
r2 = 1 - (rss/tss)
return r2
def calc_corr (y, y_pred):
num = 0
dem1 = 0
dem2 = 0
for i in range(len(y)):
num = num + ((y[i]) * (y_pred[i]))
dem1 = dem1 + (y[i])**2
dem2 = dem2 + (y_pred[i])**2
num = float(num)
dem1 = math.sqrt(float(dem1))
dem2 = math.sqrt(float(dem2))
rho = num/(dem1*dem2)
return rho
def snps_intersect(list1, list2):
return list(set(list1) & set(list2))
afa_snp = "/home/pokoro/data/mesa_models/split_mesa/"+pop+"_chr"+chrom+"_genotype_chunk"+chunk+".txt"
gex = "/home/pokoro/data/mesa_models/split_mesa/"+pop+"_chr"+chrom+"_gex_chunk"+chunk+".txt"
cov_file = "/home/pokoro/data/mesa_models/"+pop+"_3_PCs.txt"
geneanotfile = "/home/pokoro/data/mesa_models/gencode.v18.annotation.parsed.txt"
snpfilepath = "/home/pokoro/data/mesa_models/"+pop+"_"+str(chrom)+"_annot.txt"
snpannot = get_filtered_snp_annot(snpfilepath)
geneannot = get_gene_annotation(geneanotfile, chrom)
cov = get_covariates(cov_file)
expr_df = get_gene_expression(gex, geneannot)
genes = list(expr_df.columns)
gt_df = get_maf_filtered_genotype(afa_snp, 0.01)
#algorithms to use
rf = RandomForestRegressor(max_depth=None, random_state=1234, n_estimators=5000)
#text file where to write out the cv results
open("/home/pokoro/data/mesa_models/python_ml_models/results/"+pop+"_rf_5000tree_check_cv_chr"+chrom+"_chunk"+chunk+".txt", "w").write("Gene_ID"+"\t"+"Gene_Name"+"\t"+"CV_R2"+"\t"+"time(s)"+"\n")
for gene in genes:
coords = get_gene_coords(geneannot, gene)
gene_name = get_gene_name(geneannot, gene)
expr_vec = expr_df[gene]
adj_exp = adjust_for_covariates(list(expr_vec), cov)
cis_gt = get_cis_genotype(gt_df, snpannot, coords)
#build the model
if (type(cis_gt) != int) & (cis_gt.shape[1] > 0):
cis_gt = cis_gt.values
#Random Forest
rf_t0 = time.time()#do rf and time it
rf_cv = str(float(mean(cross_val_score(rf, cis_gt, adj_exp.ravel(), cv=5))))
rf_t1 = time.time()
rf_tt = str(float(rf_t1 - rf_t0))
open("/home/pokoro/data/mesa_models/python_ml_models/results/"+pop+"_rf_5000tree_check_cv_chr"+chrom+"_chunk"+chunk+".txt", "a").write(gene+"\t"+gene_name+"\t"+rf_cv+"\t"+rf_tt+"\n")