-
Notifications
You must be signed in to change notification settings - Fork 1
/
batch.py
885 lines (714 loc) · 28.5 KB
/
batch.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
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
#!/usr/bin/env python3
# pylint: disable=import-error,import-outside-toplevel,missing-function-docstring,no-value-for-parameter,too-many-arguments,too-many-locals,wrong-import-order,wrong-import-position,consider-using-enumerate,chained-comparison
"""
Hail Batch workflow for the rare-variant association analysis, including:
- get relevant variants around a gene and export genotypes as plink files
- generate input files for association tests
- run association tests
"""
# import python modules
import os
import re
import sys
import click
import logging
from typing import Dict
from google.cloud import storage
from cpg_utils import to_path
from cpg_utils.hail_batch import (
copy_common_env,
dataset_path,
get_config,
init_batch,
output_path,
remote_tmpdir,
)
import numpy as np
import pandas as pd
import xarray as xr
from limix.qc import quantile_gaussianize
from scipy.stats import shapiro
import hail as hl
import hailtop.batch as hb
from cellregmap import ( # figure out how to import this from github
run_gene_set_association,
run_burden_association,
omnibus_set_association,
)
# use logging to print statements, display at info level
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s %(levelname)s %(module)s:%(lineno)d - %(message)s',
datefmt='%Y-%m-%d %H:%M:%S',
stream=sys.stderr,
)
DEFAULT_JOINT_CALL_MT = dataset_path('mt/v7.mt')
DEFAULT_ANNOTATION_HT = dataset_path(
'tob_wgs_vep/104/vep104.3_GRCh38.ht'
) # atm VEP only
CELLREGMAP_IMAGE = get_config()['workflow'][
'driver_image'
] # australia-southeast1-docker.pkg.dev/cpg-common/images/cellregmap:dev
MULTIPY_IMAGE = 'australia-southeast1-docker.pkg.dev/cpg-common/images/multipy:0.16'
# region SUBSET_VARIANTS
def filter_variants(
mt_path: str, # 'mt/v7.mt'
samples: list[str],
output_mt_path: str, # 'tob_wgs_rv/densified_rv_only.mt'
):
"""Subset hail matrix table
Input:
joint call hail matrix table
set of samples for which we have scRNA-seq data
Output:
subset hail matrix table, containing only variants that:
1) are not ref-only, 2) biallelic, 3) meet QC filters, 4) are rare (MAF<5%)
"""
# read hail matrix table object (WGS data)
init_batch()
mt = hl.read_matrix_table(mt_path)
# subset to relevant samples (samples we have scRNA-seq data for)
mt = mt.filter_cols(hl.set(samples).contains(mt.s))
# densify
mt = hl.experimental.densify(mt)
# filter out low quality variants and consider biallelic SNPss only (no multi-allelic, no ref-only, no indels)
mt = mt.filter_rows( # check these filters!
(hl.len(hl.or_else(mt.filters, hl.empty_set(hl.tstr))) == 0) # QC
& (hl.len(mt.alleles) == 2) # remove hom-ref
& (mt.n_unsplit_alleles == 2) # biallelic
& (hl.is_snp(mt.alleles[0], mt.alleles[1])) # SNPs
)
# filter rare variants only (MAF < 5%)
mt = hl.variant_qc(mt)
mt = mt.filter_rows(
(mt.variant_qc.AF[1] < 0.05) & (mt.variant_qc.AF[1] > 0)
| (mt.variant_qc.AF[1] > 0.95) & (mt.variant_qc.AF[1] < 1)
)
mt.write(output_mt_path, overwrite=True)
logging.info(f'Number of rare (freq<5%) and QCed biallelic SNPs: {mt.count()[0]}')
# endregion SUBSET_VARIANTS
# region GET_GENE_SPECIFIC_VARIANTS
def get_promoter_variants(
mt_path: str, # ouput path from function above
ht_path: str,
gene_details: dict[str, str], # output of make_gene_loc_dict
window_size: int,
plink_file: str, # 'tob_wgs_rv/pseudobulk_rv_association/plink_files/GENE'
):
"""Subset hail matrix table
Input:
mt_path: path to already subsetted hail matrix table
ht_path: path to VEP HT
gene_details: dict of info for current gene
window_size: int, size of flanking region around genes
plink_file: str, file prefix for writing plink data
Output:
For retained variants, that are: 1) in promoter regions and
2) within 50kb up or down-stream of the gene body (or in the gene body itself)
(on top of all filters done above)
returns nothing
"""
# read hail matrix table object (pre-filtered)
init_batch()
mt = hl.read_matrix_table(mt_path)
gene_name = gene_details['gene_name']
# get relevant chromosome
chrom = gene_details['chr']
# subset to window
# get gene body position (start and end) and build interval
left_boundary = max(1, int(gene_details['start']) - window_size)
right_boundary = min(
int(gene_details['end']) + window_size,
hl.get_reference('GRCh38').lengths[chrom],
)
# get gene-specific genomic interval
gene_interval = f'{chrom}:{left_boundary}-{right_boundary}'
logging.info(f'Interval considered: {gene_interval}') # 'chr22:23219960-23348287'
# include variants up to {window size} up- and downstream
mt = hl.filter_intervals(
mt, [hl.parse_locus_interval(gene_interval, reference_genome='GRCh38')]
)
mt_path = output_path(f'{gene_name}_in_window.mt', 'tmp')
mt = mt.checkpoint(
mt_path, overwrite=True
) # add checkpoint to avoid repeat evaluation
logging.info(f'Number of variants within interval: {mt.count()[0]}')
# annotate using VEP
vep_ht = hl.read_table(ht_path)
mt = mt.annotate_rows(vep=vep_ht[mt.row_key].vep)
# filter variants found to be in promoter regions
mt = mt.filter_rows(
mt.vep.regulatory_feature_consequences['biotype'].contains('promoter')
)
promoter_path = output_path(f'{gene_name}promoter_variants.mt', 'tmp')
mt = mt.checkpoint(promoter_path, overwrite=True) # checkpoint
logging.info(
f'Number of rare (freq<5%) QC-passing, biallelic SNPs in promoter regions: {mt.count()[0]}'
)
# export this as a Hail table for downstream analysis
ht_path = output_path(
f'summary_hts/{gene_name}_rare_promoter_summary.ht', 'analysis'
)
ht = mt.rows()
ht.write(ht_path, overwrite=True)
# export MT object to PLINK (promoter variants)
# pylint: disable=import-outside-toplevel
from hail.methods import export_plink
export_plink(mt, plink_file, ind_id=mt.s)
# endregion GET_GENE_SPECIFIC_VARIANTS
# region PREPARE_INPUT_FILES
def prepare_input_files(
gene_name: str,
cell_type: str,
genotype_file_bed: str,
genotype_file_bim: str,
genotype_file_fam: str,
phenotype_file: str,
kinship_file: str,
sample_mapping_file: str,
):
"""Prepare association test input files
Input:
genotype: plink files
phenotype: gene expression
Output:
genotype matrix
phenotype vector
"""
# check that variants exist for this gene, before importing
bim_file = to_path(genotype_file_bim)
if (not bim_file.exists()) or bim_file.stat().st_size == 0:
logging.info(f'{bim_file} does not exist')
return None
from pandas_plink import read_plink1_bin
expression_filename = to_path(
output_path(f'expression_files/{gene_name}_{cell_type}.csv')
)
genotype_filename = to_path(
output_path(f'genotype_files/{gene_name}_rare_regulatory.csv')
)
kinship_filename = to_path(output_path(f'{gene_name}_kinship_common_samples.csv'))
# read in phenotype file (tsv)
phenotype = pd.read_csv(phenotype_file, sep='\t', index_col=0)
phenotype = xr.DataArray(
phenotype.values,
dims=['sample', 'gene'],
coords={'sample': phenotype.index.values, 'gene': phenotype.columns.values},
)
# read in genotype file (plink format)
to_path(genotype_file_bed).copy('temp.bed') # bed
bim_file.copy('temp.bim') # bim
to_path(genotype_file_fam).copy('temp.fam') # fam
geno = read_plink1_bin('temp.bed')
if kinship_file is not None:
# read in GRM (genotype relationship matrix; kinship matrix)
kinship = pd.read_csv(kinship_file, index_col=0)
kinship.index = kinship.index.astype('str')
assert all(
kinship.columns == kinship.index
) # symmetric matrix, donors x donors
kinship = xr.DataArray(
kinship.values,
dims=['sample_0', 'sample_1'],
coords={'sample_0': kinship.columns, 'sample_1': kinship.index},
)
kinship = kinship.sortby('sample_0').sortby('sample_1')
# this file will map different IDs (and OneK1K ID to CPG ID)
sample_mapping = pd.read_csv(dataset_path(sample_mapping_file), sep='\t')
# ensure samples are the same and in the same order across input files
# samples with expression data
donors_exprs = set(phenotype.sample.values).intersection(
set(sample_mapping['OneK1K_ID'].unique())
)
logging.info(f'Number of unique donors with expression data: {len(donors_exprs)}')
# samples with genotype data
donors_geno = set(geno.sample.values).intersection(
set(sample_mapping['InternalID'].unique())
)
logging.info(f'Number of unique donors with genotype data: {len(donors_geno)}')
# samples with both (can this be done in one step?)
sample_mapping1 = sample_mapping.loc[sample_mapping['OneK1K_ID'].isin(donors_exprs)]
sample_mapping_both = sample_mapping1.loc[
sample_mapping1['InternalID'].isin(donors_geno)
]
donors_e = sample_mapping_both['OneK1K_ID'].unique()
donors_g = sample_mapping_both['InternalID'].unique()
assert len(donors_e) == len(donors_g)
if kinship_file is not None:
# samples in kinship
donors_e_short = [re.sub('.*_', '', donor) for donor in donors_e]
donors_k = sorted(
set(list(kinship.sample_0.values)).intersection(donors_e_short)
)
logging.info(f'Number of unique common donors: {len(donors_g)}')
# subset files
# phenotype
phenotype = phenotype.sel(sample=donors_e)
# select gene
y = phenotype.sel(gene=gene_name)
y = quantile_gaussianize(y)
del phenotype # delete to free up memory
# make data frame to save as csv
y_df = pd.DataFrame(
data=y.values.reshape(y.shape[0], 1), index=y.sample.values, columns=[gene_name]
)
# genotype
geno = geno.sel(sample=donors_g)
# make data frame to save as csv
data = geno.values
geno_df = pd.DataFrame(data, columns=geno.snp.values, index=geno.sample.values)
geno_df = geno_df.dropna(axis=1)
# delete large files to free up memory
del geno
if kinship_file is not None:
# kinship
kinship = kinship.sel(sample_0=donors_k, sample_1=donors_k)
assert all(kinship.sample_0 == donors_k)
assert all(kinship.sample_1 == donors_k)
# make data frame to save as csv
kinship_df = pd.DataFrame(
kinship.values, columns=kinship.sample_0, index=kinship.sample_1
)
del kinship # delete kinship to free up memory
# save files
with expression_filename.open('w') as ef:
y_df.to_csv(ef, index=False)
with genotype_filename.open('w') as gf:
geno_df.to_csv(gf, index=False)
if kinship_file is not None:
with kinship_filename.open('w') as kf:
kinship_df.to_csv(kf, index=False)
else:
kinship_df = None
return y_df, geno_df, kinship_df
# endregion PREPARE_INPUT_FILES
# region GET_CRM_PVALUES
def get_crm_pvs(pheno, covs, genotypes, contexts=None):
"""
CellRegMap-RV tests
* score test (variance)
* burden test (max, sum, comphet)
* omnibus (Cauchy) test
Input:
input files
Output:
list of p-values from the three tests
"""
pv_norm = shapiro(pheno).pvalue
pv0 = run_gene_set_association(y=pheno, G=genotypes, W=covs, E=contexts)[0]
pv1 = run_burden_association(
y=pheno, G=genotypes, W=covs, E=contexts, mask='mask.max'
)[0]
pv2 = run_burden_association(
y=pheno, G=genotypes, W=covs, E=contexts, mask='mask.sum'
)[0]
pv3 = run_burden_association(
y=pheno, G=genotypes, W=covs, E=contexts, mask='mask.comphet'
)[0]
pv4 = omnibus_set_association(np.array([pv0, pv1]))
pv5 = omnibus_set_association(np.array([pv0, pv2]))
pv6 = omnibus_set_association(np.array([pv0, pv3]))
return np.array([pv_norm, pv0, pv1, pv2, pv3, pv4, pv5, pv6])
# endregion GET_CRM_PVALUES
# region RUN_ASSOCIATION
def run_gene_association(
gene_name: str, # 'VPREB3'
prepared_inputs: hb.resource.PythonResult,
pv_path: str, # 'Bnaive/VPREB3_pvals.csv'
):
"""Run gene-set association test
Input:
input files (genotype, phenotype)
* already in matrix / vector form
* only matching samples, correct irder
Output:
table with p-values
"""
# if the previous method depdency returned None, we will fail to
# unpack dataframes from it
if prepared_inputs is None:
return prepared_inputs
from numpy import eye, ones
# read the 3 dataframes generated by the previous job
p_df, g_df, _ = prepared_inputs
# because the current matrix is counting the copies of the reference allele
# while we are interested in the alternative allele, flip the genotypes
genotypes = 2 - g_df
# get phenptypes
pheno = p_df.values
# covariates (intercept at present)
covs = ones((genotypes.shape[0], 1)) # intercept of ones as covariates
# contexts (no context-specific analysis now, just identity)
contexts = eye(genotypes.shape[0])
# TODO: kinship (not at this stage)
cols = np.array(
[
'P_shapiro',
'P_CRM_VC',
'P_CRM_burden_max',
'P_CRM_burden_sum',
'P_CRM_burden_comphet',
'P_CRM_omnibus_max',
'P_CRM_omnibus_sum',
'P_CRM_omnibus_comphet',
]
)
# create p-values data frame
pvalues = get_crm_pvs(pheno, covs, genotypes, contexts)
pv_df = pd.DataFrame(
data=pvalues.reshape(pvalues.shape[0], 1).T,
columns=cols,
index=[gene_name],
)
pv_filename = to_path(pv_path)
with pv_filename.open('w') as pf:
pv_df.to_csv(pf)
return str(pv_filename)
# endregion RUN_ASSOCIATION
# region AGGREGATE_RESULTS
def summarise_association_results(
# pv_dfs: list[str],
celltype: str,
pv_all_filename_str: str,
):
"""Summarise results
Input:
p-values from all association tests
Output:
one csv table per cell type,
combining results across all genes in a single file
"""
from multipy.fdr import qvalue
logging.info('before glob (pv files) - summarise job')
storage_client = storage.Client()
bucket = get_config()['storage']['default']['default'].removeprefix('gs://')
prefix = f"{get_config()['workflow']['output_prefix']}/{celltype}/"
existing_pv_files = set(
f'gs://{bucket}/{filepath.name}'
for filepath in storage_client.list_blobs(bucket, prefix=prefix, delimiter='/')
if filepath.name.endswith('_pvals.csv')
)
logging.info(f'after glob - {len(existing_pv_files)} pv files to summarise')
if len(existing_pv_files) == 0:
raise Exception('No PV files, nothing to do')
pv_all_df = pd.concat(
[pd.read_csv(to_path(pv_df), index_col=0) for pv_df in existing_pv_files]
)
# run qvalues for all tests (multiple testing correction)
_, qvals = qvalue(pv_all_df['P_CRM_VC'])
pv_all_df['Q_CRM_VC'] = list(qvals)
_, qvals = qvalue(pv_all_df['P_CRM_burden_max'])
pv_all_df['Q_CRM_burden_max'] = list(qvals)
_, qvals = qvalue(pv_all_df['P_CRM_burden_sum'])
pv_all_df['Q_CRM_burden_sum'] = list(qvals)
_, qvals = qvalue(pv_all_df['P_CRM_burden_comphet'])
pv_all_df['Q_CRM_burden_comphet'] = list(qvals)
_, qvals = qvalue(pv_all_df['P_CRM_omnibus_max'])
pv_all_df['Q_CRM_omnibus_max'] = list(qvals)
_, qvals = qvalue(pv_all_df['P_CRM_omnibus_sum'])
pv_all_df['Q_CRM_omnibus_sum'] = list(qvals)
_, qvals = qvalue(pv_all_df['P_CRM_omnibus_comphet'])
pv_all_df['Q_CRM_omnibus_comphet'] = list(qvals)
pv_all_filename = to_path(pv_all_filename_str)
logging.info(f'Write summary results to {pv_all_filename}')
with pv_all_filename.open('w') as pf:
pv_all_df.to_csv(pf)
# endregion AGGREGATE_RESULTS
# region MISCELLANEOUS
def make_gene_loc_dict(file) -> dict[str, dict]:
"""
Turn gene information into a dictionary
to avoid opening this file for every gene
"""
from csv import DictReader
gene_dict = {}
with open(to_path(file)) as handle:
reader = DictReader(handle, delimiter='\t')
for row in reader:
gene_dict[row['gene_name']] = row
return gene_dict
# can probably be merged with below
def extract_genes(gene_list, expression_tsv_path) -> list[str]:
"""
Takes a list of all genes and subsets to only those
present in the expression file of interest
"""
expression_df = pd.read_csv(to_path(expression_tsv_path), sep='\t')
expression_df = filter_lowly_expressed_genes(expression_df)
gene_ids = set(list(expression_df.columns.values)[1:])
genes = set(gene_list).intersection(gene_ids)
logging.info(f'Total genes to run: {len(list(sorted(genes)))}')
return list(sorted(genes))
# copied from https://github.com/populationgenomics/tob-wgs/blob/main/scripts/eqtl_hail_batch/launch_eqtl_spearman.py
# generalised to specify min pct samples as input
def filter_lowly_expressed_genes(expression_df, min_pct=10):
"""Remove genes with low expression in all samples
Input:
expression_df: a data frame with samples as rows and genes as columns,
containing normalised expression values (i.e., the average number of molecules
for each gene detected in each person).
Returns:
A filtered version of the input data frame, after removing columns (genes)
with 0 values in more than {min_pct}% of the rows (samples) - 10 by default.
"""
# Remove genes with 0 expression in all samples
expression_df = expression_df.loc[:, (expression_df != 0).any(axis=0)]
genes_not_equal_zero = expression_df.iloc[:, 1:].values != 0
n_expr_over_zero = pd.DataFrame(genes_not_equal_zero.sum(axis=0))
percent_expr_over_zero = (n_expr_over_zero / len(expression_df.index)) * 100
percent_expr_over_zero.index = expression_df.columns[1:]
# Filter genes with less than 10 percent individuals with non-zero expression
atleastNpercent = percent_expr_over_zero[(percent_expr_over_zero > min_pct)[0]]
sample_ids = expression_df['sampleid']
expression_df = expression_df[atleastNpercent.index]
expression_df.insert(loc=0, column='sampleid', value=sample_ids)
return expression_df
def remove_sc_outliers(df, outliers=None):
"""
Remove outlier samples, as identified by single-cell analysis
"""
if outliers is None:
outliers = ['966_967', '88_88']
else:
outliers = outliers.extend(['966_967', '88_88'])
df = df[-df['OneK1K_ID'].isin(outliers)]
return df
# endregion MISCELLANEOUS
config = get_config()
@click.command()
@click.option('--celltypes')
@click.option('--expression-files-prefix', default='scrna-seq/grch38_association_files')
@click.option(
'--sample-mapping-file-tsv',
default='scrna-seq/grch38_association_files/OneK1K_CPG_IDs.tsv',
)
@click.option('--mt-path', default=DEFAULT_JOINT_CALL_MT)
@click.option('--anno-ht-path', default=DEFAULT_ANNOTATION_HT)
@click.option(
'--chromosomes',
help='List of chromosome numbers to run rare variant association analysis on. '
'Space separated, as one argument (Default: all)',
)
@click.option('--genes', default=None)
@click.option(
'--max-gene-concurrency',
type=int,
default=50,
help=(
'To avoid resource starvation, set this concurrency to limit horizontal scale. '
'Higher numbers have a better walltime, but risk jobs that are stuck (which are expensive)'
),
)
def crm_pipeline(
celltypes: str,
expression_files_prefix: str,
sample_mapping_file_tsv: str,
mt_path: str,
anno_ht_path: str,
chromosomes: str = 'all',
genes: str | None = None,
window_size: int = 50000,
max_gene_concurrency=100,
):
sb = hb.ServiceBackend(
billing_project=get_config()['hail']['billing_project'],
remote_tmpdir=remote_tmpdir(),
)
batch = hb.Batch('CellRegMap pipeline', backend=sb)
# extract samples for which we have single-cell (sc) data
sample_mapping_file = pd.read_csv(dataset_path(sample_mapping_file_tsv), sep='\t')
sample_mapping_file = remove_sc_outliers(sample_mapping_file)
sc_samples = sample_mapping_file['InternalID'].unique()
# filter to QC-passing, rare, biallelic variants
output_mt_path = output_path('densified_rv_only.mt')
if not to_path(output_mt_path).exists():
filter_job = batch.new_python_job(name='MT filter job')
copy_common_env(filter_job)
filter_job.image(CELLREGMAP_IMAGE)
filter_job.call(
filter_variants,
mt_path=mt_path,
samples=list(sc_samples),
output_mt_path=output_mt_path,
)
else:
logging.info('File already exists no need to filter')
filter_job = None
# grab all relevant genes across all chromosomes
# simpler if gene details are condensed to one file
gene_dict: dict[str, dict] = {}
if chromosomes == 'all':
# autosomes only for now
chromosomes_list = list(np.arange(22) + 1)
else:
chromosomes_list = chromosomes.split(' ')
for chromosome in chromosomes_list:
geneloc_tsv_path = dataset_path(
os.path.join(
expression_files_prefix,
'gene_location_files',
f'GRCh38_geneloc_chr{chromosome}.tsv',
)
)
# concatenating across chromosomes to have a single dict
gene_dict.update(make_gene_loc_dict(geneloc_tsv_path))
# isolate to the genes we are interested in
if genes is not None:
genes_of_interest = genes.split(' ')
else:
genes_of_interest = list(gene_dict.keys())
# processing cell types (needs to be passed as a single script for click to like it)
celltype_list = celltypes.split(' ')
logging.info(f'Cell types to run: {celltype_list}')
# only run this for relevant genes
_plink_genes = set()
for celltype in celltype_list:
expression_tsv_path = dataset_path(
os.path.join(
expression_files_prefix,
'expression_files',
f'{celltype}_expression.tsv',
)
)
logging.info(f'before extracting {celltype}-expressed genes - plink files')
_plink_genes |= set(extract_genes(genes_of_interest, expression_tsv_path))
logging.info(f'after extracting {celltype}-expressed genes - plink files')
plink_genes = list(sorted(_plink_genes))
logging.info(f'Done selecting genes, total number: {len(plink_genes)}')
# Setup MAX concurrency by genes
_dependent_jobs: list[hb.job.Job] = []
def manage_concurrency_for_job(job: hb.job.Job):
"""
To avoid having too many jobs running at once, we have to limit concurrency.
"""
if len(_dependent_jobs) >= max_gene_concurrency:
job.depends_on(_dependent_jobs[-max_gene_concurrency])
_dependent_jobs.append(job)
# for each gene, extract relevant variants (in window + with some annotation)
# submit a job for each gene (export genotypes to plink)
dependencies_dict: Dict[str, hb.job.Job] = {}
plink_root = output_path('plink_files')
logging.info('before glob (bim files)')
storage_client = storage.Client()
bucket = get_config()['storage']['default']['default'].removeprefix('gs://')
prefix = os.path.join(get_config()['workflow']['output_prefix'], 'plink_files/')
bim_files = set(
f'gs://{bucket}/{filepath.name}'
for filepath in storage_client.list_blobs(bucket, prefix=prefix, delimiter='/')
if filepath.name.endswith('bim')
)
logging.info(f'after glob: {len(bim_files)} bim files already exist')
for gene in plink_genes:
# final path for this gene - generate first (check syntax)
plink_file = os.path.join(plink_root, gene)
gene_dict[gene]['plink'] = plink_file
# if the plink output exists, do not re-generate it
if f'{plink_file}.bim' in bim_files:
continue
plink_job = batch.new_python_job(f'Create plink files for: {gene}')
manage_concurrency_for_job(plink_job)
copy_common_env(plink_job)
if filter_job:
plink_job.depends_on(filter_job)
plink_job.image(CELLREGMAP_IMAGE)
plink_job.call(
get_promoter_variants,
mt_path=output_mt_path,
ht_path=anno_ht_path,
gene_details=gene_dict[gene],
window_size=window_size,
plink_file=plink_file,
)
dependencies_dict[gene] = plink_job
# the next phase will be done for each cell type
for celltype in celltype_list:
expression_tsv_path = dataset_path(
os.path.join(
expression_files_prefix,
'expression_files',
f'{celltype}_expression.tsv',
)
)
logging.info(
f'before extracting {celltype}-expressed genes to run association for'
)
genes_list = extract_genes(genes_of_interest, expression_tsv_path)
logging.info(
f'after extracting {celltype}-expressed genes: run association for {len(genes_list)} genes'
)
# logging.info(f'Genes to run: {genes_list}')
if not genes_list:
logging.info('No genes to run, exit!')
continue
gene_run_jobs = []
# cell_type_root = output_path(celltype)
logging.info(f'before glob: pv files for {celltype}')
storage_client = storage.Client()
bucket = get_config()['storage']['default']['default'].removeprefix('gs://')
prefix = f"{get_config()['workflow']['output_prefix']}/{celltype}/"
existing_files = set(
f'gs://{bucket}/{filepath.name}'
for filepath in storage_client.list_blobs(
bucket, prefix=prefix, delimiter='/'
)
if filepath.name.endswith('_pvals.csv')
)
# existing_files = list(to_path(cell_type_root).glob('*_pvals.csv'))
logging.info(f'after glob: {len(existing_files)} pv files for {celltype}')
for gene in genes_list:
# wrapped this with output_path
pv_file = output_path(f'{celltype}/{gene}_pvals.csv')
# check if running is required
if pv_file in existing_files:
logging.info(f'We already ran associations for {gene}!')
continue
if gene_dict[gene]['plink'] is None:
logging.info(f'No plink files for {gene}, exit!')
continue
plink_output_prefix = gene_dict[gene]['plink']
# prepare input files
run_job = batch.new_python_job(f'Run association for: {celltype}, {gene}')
manage_concurrency_for_job(run_job)
copy_common_env(run_job)
if dependency := dependencies_dict.get(gene):
run_job.depends_on(dependency)
run_job.image(CELLREGMAP_IMAGE)
# the python_job.call only returns one object
# the object is a file containing y_df, geno_df, kinship_df
# all pickled into a file
input_results = run_job.call(
prepare_input_files,
gene_name=gene,
cell_type=celltype,
genotype_file_bed=plink_output_prefix + '.bed',
genotype_file_bim=plink_output_prefix + '.bim',
genotype_file_fam=plink_output_prefix + '.fam',
phenotype_file=expression_tsv_path,
kinship_file=None, # change this to work when this file is needed
sample_mapping_file=sample_mapping_file_tsv,
)
# run association in the same python env
gene_run_jobs.append(run_job)
run_job.call(
run_gene_association,
gene_name=gene,
prepared_inputs=input_results,
pv_path=pv_file,
)
# combine all p-values across all chromosomes, genes (per cell type)
summarise_job = batch.new_python_job(f'Summarise all results for {celltype}')
copy_common_env(summarise_job)
summarise_job.depends_on(*gene_run_jobs)
summarise_job.image(MULTIPY_IMAGE)
pv_all_filename_csv = str(
output_path(f'{celltype}_all_pvalues.csv', 'analysis')
)
summarise_job.call(
summarise_association_results,
celltype=celltype,
pv_all_filename_str=str(pv_all_filename_csv),
)
# set jobs running
batch.run(wait=False)
if __name__ == '__main__':
crm_pipeline()