-
Notifications
You must be signed in to change notification settings - Fork 0
/
mesa_svr.py
491 lines (356 loc) · 16.2 KB
/
mesa_svr.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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
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
#MESA
cis_gt = pd.read_csv("/Users/okoro/OneDrive/Desktop/svr/svr_cis_gt_chr6_HLA-DRB1.csv")
adj_exp = pd.read_csv("/Users/okoro/OneDrive/Desktop/svr/svr_adj_expression_chr6_HLA-DRB1.csv")
#convert the dataframe to numpy array
cis_gt = cis_gt.values
adj_exp = adj_exp.values
#split data into train and test
#X_train, X_test, y_train, y_test = train_test_split(cis_gt, adj_exp, test_size=0.1, random_state=42)
#print()
print("Random Forest")
#run a random forest regression
from sklearn.ensemble import RandomForestRegressor
#CV
rf = RandomForestRegressor(max_depth=None, random_state=1234, n_estimators=100)
t0 = time.time() # or time.process_time()
print(mean(cross_val_score(rf, cis_gt, adj_exp.ravel(), cv=5)))
t1 = time.time() #or time.process_time()
total_time = t1 - t0
#fit the rf with all data
rf.fit(cis_gt, adj_exp.ravel())
std = np.std(rf.feature_importances_)
importances = rf.feature_importances_
indices = np.argsort(importances)[::-1]
print("Feature ranking:")
for f in range(cis_gt.shape[1]):
print("%d. feature %d (%f)" % (f + 1, indices[f], importances[indices[f]]))
#select only the most important features (SNPs)
imp_indices = indices[0:500] #first 500 most imprtant indices
imp_cis_gt = cis_gt[:,imp_indices]
#Try simple linear regression with the feature importance
from sklearn import linear_model
reg = linear_model.LinearRegression() #it performed very poor
mean(cross_val_score(reg, cis_gt, adj_exp, cv=5))
mean(cross_val_score(reg, imp_cis_gt, adj_exp, cv=5)
from sklearn.svm import SVR
svrl = SVR(kernel="linear", gamma="auto")
mean(cross_val_score(svrl, cis_gt, adj_exp.ravel(), cv=5))
mean(cross_val_score(svrl, imp_cis_gt, adj_exp.ravel(), cv=5))
svr = SVR(kernel="rbf", gamma="auto")
mean(cross_val_score(svr, cis_gt, adj_exp.ravel(), cv=5))
mean(cross_val_score(svr, imp_cis_gt, adj_exp.ravel(), cv=5))
from sklearn.neighbors import KNeighborsRegressor
knn = KNeighborsRegressor(n_neighbors=10, weights = "distance")
mean(cross_val_score(knn, cis_gt, adj_exp, cv=5))
mean(cross_val_score(knn, imp_cis_gt, adj_exp, cv=5))
from sklearn.linear_model import ElasticNet
elnet = ElasticNet(alpha=0.1, random_state=1234)
mean(cross_val_score(elnet, cis_gt, adj_exp, cv=5))
mean(cross_val_score(elnet, imp_cis_gt, adj_exp, cv=5))
#KNN performed the best with feature importance R2 0.89
#followed by elasticNet R2 0.78
#svm performd at R2 0.53
imp_snps = cis_gt.columns[imp_indices]
pruned_cis_gt = cis_gt[imp_snps]
pruned_cis_gt = pruned_cis_gt.values
#this is how to index the pandas dataframe to get the snp names of the
# important features after running random forest feature importance
# this will be done so as to get the snp names, and then look for those snp
# names in the test dataframe. then take the overlaps and create a new test
# dataframe. Then fit the ML algorithm with the old important features and then
# predict expression in the new pruned test set
#to get column names of pandas dataframe
# for col in cis_gt.columns:
# print(col)
# to access specific column name by col index e.g cis_gt.columns[1]
# to access list of col names by col indices e.g cis_gt.columns[indices]
# to the get the important colnames as selected by RF e.g cis_gt.columns[imp_indices]
# to finally create or select out only important features columns
# is df[colname] or df[list of colnames] e.g cis_gt[cis_gt.columns[imp_indices]]
# store the pruned cis_dataframe in a new variable
# e.g pruned_cis = cis_gt[cis_gt.columns[imp_indices]]
#then fit your ML with the pruned cis
#note however, that your measured expression should be kept intact
#use these same steps to select out the important snps from the test data
#then use the fitted ML model to predict expression on the overlap snps of test
#data. then do the correlation of the
#make sure that the imp_snps are also present in the test snps
#therefore take the intersect of the two snps lists
# example function to take intersection of two lists
def snps_intersect(list1, list2):
return list(set(list1) & set(list2))
def snps_intersect(list1, list2):
return set(list1).intersection(list2)
#take snps intersect
snp_intersect = snps_intersect(imp_snps, yri_snps)
pruned_test = yri_cis[snp_intersect]
pruned_test = pruned_test.values
pruned_train = afa_cis[snp_intersect]
pruned_train = pruned_train.values
mean(cross_val_score(rf, pruned_train, afaadj_np.ravel(), cv=5))
rf.fit(pruned_train, afaadj_np.ravel())
rf.score(pruned_test, yriadj_np.ravel())
ypred = rf.predict(pruned_test)
calc_corr(yriadj_np, ypred)
#turn all these code and data file into a wrapper, and or a docker file that can be installed and run by users... with database capability
#even better, without having to save the data, save the fitted model objects for each algorithm into a file, and reload from there.
#this object savings can be done with python pickle module. import pickle
#test in YRI
test_adj = pd.read_csv("/Users/okoro/OneDrive/Desktop/svr/svr_adj_expression_chr1_CRYZ_YRI.csv")
test_cis = pd.read_csv("/Users/okoro/OneDrive/Desktop/svr/svr_cis_gt_chr1_CRYZ_YRI.csv")
with open("/Users/okoro/OneDrive/Desktop/svr/elastic", "wb") as el:
pickle.dump(elastic, el)
nu_el = None
with open("/Users/okoro/OneDrive/Desktop/svr/elastic", "rb") as nu:
nu_el = pickle.load(nu)
import math
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
snpfilepath = "/Users/okoro/OneDrive/Desktop/svr/AFA_1_annot.txt"
snpfile = "/Users/okoro/OneDrive/Desktop/svr/YRI_annot.chr1.txt"
#snpfile = "/Users/okoro/OneDrive/Desktop/svr/yri_dummy_anot.txt"
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
geneanotfile = "/Users/okoro/OneDrive/Desktop/svr/gencode.v18.annotation.parsed.txt"
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
# gene has to be string
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
cov_file = "/Users/okoro/OneDrive/desktop/svr/AFA_3_PCs.txt"
yri_covfile = "/Users/okoro/OneDrive/Desktop/svr/YRI_final_pcs.txt"
#cov file is space separated " ".
# the column names contain IID which is sample ID
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
gex = "/Users/okoro/OneDrive/Desktop/svr/meqtl_sorted_AFA_MESA_Epi_GEX_data_sidno_Nk-10.txt"
gex_yri = "/Users/okoro/OneDrive/Desktop/svr/YRI_expression_ens.txt"
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
#newdf = DataFrame(scale(df), index=df.index, columns=df.columns) # if we want
#to retain the scale array to dataframe with rownames and colnames
# yri_exp["ENSG00000116791.9"] # test gene expression vector
# pproblem gene in chr22 ENSG00000117215.10
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
yrisnpfile = "/Users/okoro/OneDrive/Desktop/svr/YRI_snp.chr1.txt"
afa_snp = "/Users/okoro/OneDrive/Desktop/svr/AFA_1_snp.txt"
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 NaN
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
#Still working on this
"""
do_covariance <- function(gene_id, cis_gt, rsids, varIDs) {
model_gt <- cis_gt[,varIDs, drop=FALSE]
colnames(model_gt) <- rsids
geno_cov <- cov(model_gt)
geno_cov[lower.tri(geno_cov)] <- NA
cov_df <- melt(geno_cov, varnames = c("rsid1", "rsid2"), na.rm = TRUE) %>%
mutate(gene=gene_id) %>%
select(GENE=gene, RSID1=rsid1, RSID2=rsid2, VALUE=value) %>%
arrange(GENE, RSID1, RSID2)
cov_df
}
def do_covariance (gene_id, cis_gt, rsids, varIDs): #working on this but have not found solution
model_gt = cis_gt[varIDs]
model.columns = rsids
geno_cov = model_gt.values
geno_cov = np.cov(geno_cov) #you can join to do, but i will pause here.
"""
cour = [[70,90,80],
[60,90,70],
[40,55,50],
[60,40,50],
[70,70,75],
[70,50,40],
[30,40,39],
[50,40,45],
[60,40,45],
[35,40,50]]
iq = [90, 92, 65, 50, 86, 61, 40, 45, 50, 49]
#trying to do the covariance
sv = list(snpannot["varID"])
varid = snps_intersect(cc,sv) #cc is cis_gt columns
sn = snpannot[(snpannot.varID.isin(varid))]
#Elastic Net implementation equivalence in from R
elnet = ElasticNet(alpha=0.1, random_state=1234)
x = cis_gt.values
y = adj_exp
elnet.fit(x,y)
beta = elnet.coef_ #read in the beta coefficients of the
indices2 = np.where(beta > 0) #find the indices of the positive betas
pos_beta = beta[indices2] #find the beta values themselves
weighted_snps = cis_gt.columns[indices2] #find the snps with the positive betas
#filter the snpannot by the weighted snps
sn = snpannot[(snpannot.varID.isin(weighted_snps))]
#tyring to do covariance
weighted_varID = list(sn.varID)
weighted_rsid = list(sn.rsid)
model_gt = cis_gt[weighted_varID]
model_gt.columns = weighted_rsid
#implement how the get the weight files for elastic net
#columns are:
# gene_id rsid varID ref alt beta
#use loop to write out the weights and access the snpannot info
#first drop the chr and pos colums to minimize size to 4 columns
# varID refAllele effectAllele rsid
sn.drop(labels=['chr','pos'], axis='columns', inplace=True)
#loop and keep everything in a sorted order
#this can be use for the covariance function
gene_id = []
rsid = []
varID = []
ref = []
alt = []
weights = []
for i in range(len(pos_beta)):
gene_id.append(gene)#add the gene_id
weights.append(pos_beta[i])#add the snp weight
snpid = weighted_snps[i]#take put snp id
sna = sn[sn.varID==snpid] #filter snp annot by snpid
rsid.append(sna.iloc[0,3]) #take out snp rsid
varID.append(sna.iloc[0,0]) #take out snp varID
ref.append(sna.iloc[0,1]) #take out snp refAllele
alt.append(sna.iloc[0,2]) #take out snp effectAllele
#use the lists for the covariance function
def do_covariance(gene,cis_gt,rsid,varID): #except gene_id, the arguments are list
model_gt = cis_gt[varID]
model_gt.columns = rsid
gcov = np.cov(model_gt, rowvar=False) #If rowvar is True (default), then each row represents a variable, with observations in the columns. which is not true. so rowvar=False
gcov = pd.DataFrame(gcov)
gcov.columns = rsid
gcov.index = rsid
for i in range(len(gcov)): #Best looper for the covariance
for j in range(i, len(gcov)):
print(gene, gcov.index[i], gcov.columns[j], gcov.iloc[i,j])
#OR
#convert the lists to pandas dataframe
#use numpy
w_file = pd.DataFrame(np.column_stack([gene_id, rsid, varID, ref, alt, weights]),
columns=['gene_id', 'rsid', 'varID', 'ref', 'alt', 'beta'])
#or use dictionary... however this truncates the decimal places
w_file2 = pd.DataFrame({'gene_id':gene_id, 'rsid':rsid, 'varID':varID,
'ref':ref, 'alt':alt, 'beta':weights})
#instead of puting the file in list..., write it out to file
open("/Users/okoro/OneDrive/Desktop/weights.txt",
"w").write("gene_id"+"\t"+"rsid"+"\t"+"varID"+"\t"+"ref"+"\t"+"alt"+"\t"+"beta"+"\n")
for i in range(len(pos_beta)):
#add the gene_id
w = str(pos_beta[i])#add the snp weight
snpid = weighted_snps[i]#take put snp id
sna = sn[sn.varID==snpid] #filter snp annot by snpid
rs = str(sna.iloc[0,3]) #take out snp rsid
va = str(sna.iloc[0,0]) #take out snp varID
re = str(sna.iloc[0,1]) #take out snp refAllele
al = str(sna.iloc[0,2]) #take out snp effectAllele
open("/Users/okoro/OneDrive/Desktop/weights.txt",
"a").write(gene+"\t"+rs+"\t"+va+"\t"+re+"\t"+al+"\t"+w+"\n")
i = 0
while i < len(gc): #wrong looper
for j in range(len(gc)):
print(gc.columns[i], gc.index[j-i], gc.iloc[j-i,i])
i+=1
#simulate the covariance file and try melt
col1 = [3,4,7]
col2 = [4,5,7]
col2 = [4,5,6]
col3 = [7,6,8]
eg = pd.DataFrame({'rs01':col1, 'rs02':col2, 'rs03':col3})
row = ['rs01', 'rs02', 'rs03']
eg.index = row
i=0
while i < len(eg):#wrong looper
for j in range(len(eg)):
print(eg.index[i], eg.columns[j-i], eg.iloc[j-i,i])
i+=1
i = 0
j = 0
k = 0
while i < len(eg):#wrong looper
while k < len(eg):
print(eg.index[i], eg.columns[j-i], eg.iloc[j-i,i])
k+=1
i+=1
j = i+1
for i in range(eg.shape[0]): #good too
for j in range(i, eg.shape[1]):
print(eg.index[i], eg.columns[j], eg.iloc[i,j])
for i in range(len(eg)): #Best looper for the covariance
for j in range(i, len(eg)):
print(eg.index[i], eg.columns[j], eg.iloc[i,j])