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 | @dataclass
class FinnGenFinemapping:
"""SuSIE finemapping dataset for FinnGen.
Credible sets from SuSIE are extracted and transformed into StudyLocus objects:
- Study ID in the special format (e.g. FINNGEN_R11*)
- Credible set specific finemapping statistics (e.g. LogBayesFactors, Alphas/Posterior)
- Additional credible set level BayesFactor filtering is applied (LBF > 2)
- StudyLocusId is annotated for each credible set.
Finemapping method is populated as a constant ("SuSIE").
"""
raw_schema: t.StructType = StructType(
[
StructField("trait", StringType(), True),
StructField("region", StringType(), True),
StructField("v", StringType(), True),
StructField("rsid", StringType(), True),
StructField("chromosome", StringType(), True),
StructField("position", StringType(), True),
StructField("allele1", StringType(), True),
StructField("allele2", StringType(), True),
StructField("maf", StringType(), True),
StructField("beta", StringType(), True),
StructField("se", StringType(), True),
StructField("p", StringType(), True),
StructField("mean", StringType(), True),
StructField("sd", StringType(), True),
StructField("prob", StringType(), True),
StructField("cs", StringType(), True),
StructField("cs_specific_prob", DoubleType(), True),
StructField("low_purity", StringType(), True),
StructField("lead_r2", StringType(), True),
StructField("mean_99", StringType(), True),
StructField("sd_99", StringType(), True),
StructField("prob_99", StringType(), True),
StructField("cs_99", StringType(), True),
StructField("cs_specific_prob_99", StringType(), True),
StructField("low_purity_99", StringType(), True),
StructField("lead_r2_99", StringType(), True),
StructField("alpha1", DoubleType(), True),
StructField("alpha2", DoubleType(), True),
StructField("alpha3", DoubleType(), True),
StructField("alpha4", DoubleType(), True),
StructField("alpha5", DoubleType(), True),
StructField("alpha6", DoubleType(), True),
StructField("alpha7", DoubleType(), True),
StructField("alpha8", DoubleType(), True),
StructField("alpha9", DoubleType(), True),
StructField("alpha10", DoubleType(), True),
StructField("mean1", StringType(), True),
StructField("mean2", StringType(), True),
StructField("mean3", StringType(), True),
StructField("mean4", StringType(), True),
StructField("mean5", StringType(), True),
StructField("mean6", StringType(), True),
StructField("mean7", StringType(), True),
StructField("mean8", StringType(), True),
StructField("mean9", StringType(), True),
StructField("mean10", StringType(), True),
StructField("sd1", StringType(), True),
StructField("sd2", StringType(), True),
StructField("sd3", StringType(), True),
StructField("sd4", StringType(), True),
StructField("sd5", StringType(), True),
StructField("sd6", StringType(), True),
StructField("sd7", StringType(), True),
StructField("sd8", StringType(), True),
StructField("sd9", StringType(), True),
StructField("sd10", StringType(), True),
StructField("lbf_variable1", DoubleType(), True),
StructField("lbf_variable2", DoubleType(), True),
StructField("lbf_variable3", DoubleType(), True),
StructField("lbf_variable4", DoubleType(), True),
StructField("lbf_variable5", DoubleType(), True),
StructField("lbf_variable6", DoubleType(), True),
StructField("lbf_variable7", DoubleType(), True),
StructField("lbf_variable8", DoubleType(), True),
StructField("lbf_variable9", DoubleType(), True),
StructField("lbf_variable10", DoubleType(), True),
]
)
summary_schema: t.StructType = StructType(
[
StructField("trait", StringType(), True),
StructField("region", StringType(), True),
StructField("cs", StringType(), True),
StructField("cs_log10bf", DoubleType(), True),
StructField("cs_avg_r2", DoubleType(), True),
StructField("cs_min_r2", DoubleType(), True),
]
)
raw_hail_shema: hl.tstruct = hl.tstruct(
trait=hl.tstr,
region=hl.tstr,
v=hl.tstr,
rsid=hl.tstr,
chromosome=hl.tstr,
position=hl.tstr,
allele1=hl.tstr,
allele2=hl.tstr,
maf=hl.tstr,
beta=hl.tstr,
se=hl.tstr,
p=hl.tstr,
mean=hl.tstr,
sd=hl.tstr,
prob=hl.tstr,
cs=hl.tstr,
cs_specific_prob=hl.tfloat64,
low_purity=hl.tstr,
lead_r2=hl.tstr,
mean_99=hl.tstr,
sd_99=hl.tstr,
prob_99=hl.tstr,
cs_99=hl.tstr,
cs_specific_prob_99=hl.tstr,
low_purity_99=hl.tstr,
lead_r2_99=hl.tstr,
alpha1=hl.tfloat64,
alpha2=hl.tfloat64,
alpha3=hl.tfloat64,
alpha4=hl.tfloat64,
alpha5=hl.tfloat64,
alpha6=hl.tfloat64,
alpha7=hl.tfloat64,
alpha8=hl.tfloat64,
alpha9=hl.tfloat64,
alpha10=hl.tfloat64,
mean1=hl.tstr,
mean2=hl.tstr,
mean3=hl.tstr,
mean4=hl.tstr,
mean5=hl.tstr,
mean6=hl.tstr,
mean7=hl.tstr,
mean8=hl.tstr,
mean9=hl.tstr,
mean10=hl.tstr,
sd1=hl.tstr,
sd2=hl.tstr,
sd3=hl.tstr,
sd4=hl.tstr,
sd5=hl.tstr,
sd6=hl.tstr,
sd7=hl.tstr,
sd8=hl.tstr,
sd9=hl.tstr,
sd10=hl.tstr,
lbf_variable1=hl.tfloat64,
lbf_variable2=hl.tfloat64,
lbf_variable3=hl.tfloat64,
lbf_variable4=hl.tfloat64,
lbf_variable5=hl.tfloat64,
lbf_variable6=hl.tfloat64,
lbf_variable7=hl.tfloat64,
lbf_variable8=hl.tfloat64,
lbf_variable9=hl.tfloat64,
lbf_variable10=hl.tfloat64,
)
summary_hail_schema: hl.tstruct = hl.tstruct(
trait=hl.tstr,
region=hl.tstr,
cs=hl.tstr,
cs_log10bf=hl.tfloat64,
cs_avg_r2=hl.tfloat64,
cs_min_r2=hl.tfloat64,
)
@staticmethod
def _infer_block_gzip_compression(paths: str | list[str]) -> bool:
"""Naively infer compression type based on the file extension.
Args:
paths (str | list[str]): File path(s).
Returns:
bool: True if block gzipped, False otherwise.
"""
if isinstance(paths, str):
return paths.endswith(".bgz")
return all(path.endswith(".bgz") for path in paths)
@classmethod
def from_finngen_susie_finemapping(
cls: type[FinnGenFinemapping],
spark: SparkSession,
finngen_susie_finemapping_snp_files: (str | list[str]),
finngen_susie_finemapping_cs_summary_files: (str | list[str]),
finngen_release_prefix: str,
credset_lbf_threshold: float = 0.8685889638065036,
) -> StudyLocus:
"""Process the SuSIE finemapping output for FinnGen studies.
The finngen_susie_finemapping_snp_files are files that contain variant summaries with credible set information with following shema:
- trait: phenotype
- region: region for which the fine-mapping was run.
- v, rsid: variant ids
- chromosome
- position
- allele1
- allele2
- maf: minor allele frequency
- beta: original marginal beta
- se: original se
- p: original p
- mean: posterior mean beta after fine-mapping
- sd: posterior standard deviation after fine-mapping.
- prob: posterior inclusion probability
- cs: credible set index within region
- lead_r2: r2 value to a lead variant (the one with maximum PIP) in a credible set
- alphax: posterior inclusion probability for the x-th single effect (x := 1..L where L is the number of single effects (causal variants) specified; default: L = 10).
- lbfx: log-Bayes Factor for each variable and single effect (i.e credible set).
- meanx: posterior mean for each variable and single effect (i.e credible set).
- sdx: posterior sd of mean for each variable and single effect (i.e credible set).
As for r11 finngen release these files are ingested from `https://console.cloud.google.com/storage/browser/finngen-public-data-r11/finemap/full/susie/` by
- *.snp.bgz
- *.snp.bgz.tbi
Each file contains index (.tbi) file that is required to read the block gzipped compressed snp file. These files needs to be
downloaded, transfromed from block gzipped to plain gzipped and then uploaded to the storage bucket, before they can be read by spark or read by hail directly as import table.
The finngen_susie_finemapping_cs_summary_files are files that Contains credible set summaries from SuSiE fine-mapping for all genome-wide significant regions with following schema:
- trait: phenotype
- region: region for which the fine-mapping was run.
- cs: running number for independent credible sets in a region, assigned to 95% PIP
- cs_log10bf: Log10 bayes factor of comparing the solution of this model (cs independent credible sets) to cs -1 credible sets
- cs_avg_r2: Average correlation R2 between variants in the credible set
- cs_min_r2: minimum r2 between variants in the credible set
- low_purity: boolean (TRUE,FALSE) indicator if the CS is low purity (low min r2)
- cs_size: how many snps does this credible set contain
- good_cs: boolean (TRUE,FALSE) indicator if this CS is considered reliable. IF this is FALSE then top variant reported for the CS will be chosen based on minimum p-value in the credible set, otherwise top variant is chosen by maximum PIP
- cs_id:
- v: top variant (chr:pos:ref:alt)
- p: top variant p-value
- beta: top variant beta
- sd: top variant standard deviation
- prob: overall PIP of the variant in the region
- cs_specific_prob: PIP of the variant in the current credible set (this and previous are typically almost identical)
- 0..n: configured annotation columns. Typical default most_severe,gene_most_severe giving consequence and gene of top variant
These files needs to be downloaded from the `https://console.cloud.google.com/storage/browser/finngen-public-data-r11/finemap/summary/` by *.cred.summary.tsv pattern,
Args:
spark (SparkSession): SparkSession object.
finngen_susie_finemapping_snp_files (str | list[str]): SuSIE finemapping output filename(s).
finngen_susie_finemapping_cs_summary_files (str | list[str]): filename of SuSIE finemapping credible set summaries.
finngen_release_prefix (str): Finngen project release prefix. Should look like FINNGEN_R*.
credset_lbf_threshold (float, optional): Filter out credible sets below, Default 0.8685889638065036 == np.log10(np.exp(2)), this is threshold from publication.
Returns:
StudyLocus: Processed SuSIE finemapping output in StudyLocus format.
"""
# NOTE: hail allows for importing block gzipped files, spark does not without external libraries.
# check https://github.com/projectglow/glow/blob/36bf6121fbc4ccc33a13b028deb87b63faeba7a9/core/src/main/scala/io/projectglow/vcf/VCFFileFormat.scala#L274
# how it could be implemented with spark.
bgzip_compressed_snps = cls._infer_block_gzip_compression(
finngen_susie_finemapping_snp_files
)
# NOTE: fallback to spark read if not block gzipped file in the input
if bgzip_compressed_snps:
snps_df = hl.import_table(
finngen_susie_finemapping_snp_files,
delimiter="\t",
types=cls.raw_hail_shema,
).to_spark()
else:
snps_df = (
spark.read.schema(cls.raw_schema)
.option("delimiter", "\t")
.option("compression", "gzip")
.csv(finngen_susie_finemapping_snp_files, header=True)
)
processed_finngen_finemapping_df = (
# Drop rows which don't have proper position.
snps_df.filter(f.col("position").cast(t.IntegerType()).isNotNull())
# Drop non credible set SNPs:
.filter(f.col("cs").cast(t.IntegerType()) > 0)
.select(
# Add study idenfitier.
f.concat_ws("_", f.lit(finngen_release_prefix), f.col("trait"))
.cast(t.StringType())
.alias("studyId"),
f.col("region"),
# Add variant information.
f.regexp_replace(f.col("v"), ":", "_").alias("variantId"),
f.col("cs").cast("integer").alias("credibleSetIndex"),
f.regexp_replace(f.col("chromosome"), "^chr", "")
.cast(t.StringType())
.alias("chromosome"),
f.col("position").cast(t.IntegerType()),
f.col("allele1").cast(t.StringType()).alias("ref"),
f.col("allele2").cast(t.StringType()).alias("alt"),
# Parse p-value into mantissa and exponent.
*parse_pvalue(f.col("p")),
# Add standard error, and allele frequency information.
f.col("se").cast("double").alias("standardError"),
f.col("maf").cast("float").alias("effectAlleleFrequencyFromSource"),
f.lit(FinemappingMethod.SUSIE.value).alias("finemappingMethod"),
*[
f.col(f"alpha{i}").cast(t.DoubleType()).alias(f"alpha_{i}")
for i in range(1, 11)
],
*[
f.col(f"lbf_variable{i}").cast(t.DoubleType()).alias(f"lbf_{i}")
for i in range(1, 11)
],
*[
f.col(f"mean{i}").cast(t.DoubleType()).alias(f"beta_{i}")
for i in range(1, 11)
],
)
.withColumn(
"posteriorProbability",
f.when(f.col("credibleSetIndex") == 1, f.col("alpha_1"))
.when(f.col("credibleSetIndex") == 2, f.col("alpha_2"))
.when(f.col("credibleSetIndex") == 3, f.col("alpha_3"))
.when(f.col("credibleSetIndex") == 4, f.col("alpha_4"))
.when(f.col("credibleSetIndex") == 5, f.col("alpha_5"))
.when(f.col("credibleSetIndex") == 6, f.col("alpha_6"))
.when(f.col("credibleSetIndex") == 7, f.col("alpha_7"))
.when(f.col("credibleSetIndex") == 8, f.col("alpha_8"))
.when(f.col("credibleSetIndex") == 9, f.col("alpha_9"))
.when(f.col("credibleSetIndex") == 10, f.col("alpha_10")),
)
.drop(
"alpha_1",
"alpha_2",
"alpha_3",
"alpha_4",
"alpha_5",
"alpha_6",
"alpha_7",
"alpha_8",
"alpha_9",
"alpha_10",
)
.withColumn(
"logBF",
f.when(f.col("credibleSetIndex") == 1, f.col("lbf_1"))
.when(f.col("credibleSetIndex") == 2, f.col("lbf_2"))
.when(f.col("credibleSetIndex") == 3, f.col("lbf_3"))
.when(f.col("credibleSetIndex") == 4, f.col("lbf_4"))
.when(f.col("credibleSetIndex") == 5, f.col("lbf_5"))
.when(f.col("credibleSetIndex") == 6, f.col("lbf_6"))
.when(f.col("credibleSetIndex") == 7, f.col("lbf_7"))
.when(f.col("credibleSetIndex") == 8, f.col("lbf_8"))
.when(f.col("credibleSetIndex") == 9, f.col("lbf_9"))
.when(f.col("credibleSetIndex") == 10, f.col("lbf_10")),
)
.drop(
"lbf_1",
"lbf_2",
"lbf_3",
"lbf_4",
"lbf_5",
"lbf_6",
"lbf_7",
"lbf_8",
"lbf_9",
"lbf_10",
)
.withColumn(
"beta",
f.when(f.col("credibleSetIndex") == 1, f.col("beta_1"))
.when(f.col("credibleSetIndex") == 2, f.col("beta_2"))
.when(f.col("credibleSetIndex") == 3, f.col("beta_3"))
.when(f.col("credibleSetIndex") == 4, f.col("beta_4"))
.when(f.col("credibleSetIndex") == 5, f.col("beta_5"))
.when(f.col("credibleSetIndex") == 6, f.col("beta_6"))
.when(f.col("credibleSetIndex") == 7, f.col("beta_7"))
.when(f.col("credibleSetIndex") == 8, f.col("beta_8"))
.when(f.col("credibleSetIndex") == 9, f.col("beta_9"))
.when(f.col("credibleSetIndex") == 10, f.col("beta_10")),
)
.drop(
"beta_1",
"beta_2",
"beta_3",
"beta_4",
"beta_5",
"beta_6",
"beta_7",
"beta_8",
"beta_9",
"beta_10",
)
)
bgzip_compressed_cs_summaries = cls._infer_block_gzip_compression(
finngen_susie_finemapping_cs_summary_files
)
# NOTE: fallback to spark read if not block gzipped file in the input
# in case we want to use the raw files from the
# https://console.cloud.google.com/storage/browser/finngen-public-data-r11/finemap/full/susie/*.cred.gz
if bgzip_compressed_cs_summaries:
cs_summary_df = hl.import_table(
finngen_susie_finemapping_cs_summary_files,
delimiter="\t",
types=cls.summary_hail_schema,
).to_spark()
else:
cs_summary_df = (
spark.read.schema(cls.summary_schema)
.option("delimiter", "\t")
.csv(finngen_susie_finemapping_cs_summary_files, header=True)
)
# drop credible sets where logbf > 2. Except when there's only one credible set in region:
# 0.8685889638065036 corresponds to np.log10(np.exp(2)), to match the orginal threshold in publication.
finngen_finemapping_summaries_df = (
# Read credible set level lbf, it is output as a different file which is not ideal.
cs_summary_df.select(
f.col("region"),
f.col("trait"),
f.col("cs").cast("integer").alias("credibleSetIndex"),
f.col("cs_log10bf").cast("double").alias("credibleSetlog10BF"),
f.col("cs_avg_r2").cast("double").alias("purityMeanR2"),
f.col("cs_min_r2").cast("double").alias("purityMinR2"),
)
.filter(
(f.col("credibleSetlog10BF") > credset_lbf_threshold)
| (f.col("credibleSetIndex") == 1)
)
.withColumn(
"studyId",
f.concat_ws("_", f.lit(finngen_release_prefix), f.col("trait")),
)
)
processed_finngen_finemapping_df = processed_finngen_finemapping_df.join(
finngen_finemapping_summaries_df,
on=["studyId", "region", "credibleSetIndex"],
how="inner",
)
toploci_df = get_top_ranked_in_window(
processed_finngen_finemapping_df,
Window.partitionBy("studyId", "region", "credibleSetIndex").orderBy(
f.desc("posteriorProbability")
),
).select(
"variantId",
"chromosome",
"position",
"studyId",
"beta",
"pValueMantissa",
"pValueExponent",
"effectAlleleFrequencyFromSource",
"standardError",
"region",
"credibleSetIndex",
"finemappingMethod",
"credibleSetlog10BF",
"purityMeanR2",
"purityMinR2",
)
processed_finngen_finemapping_df = (
processed_finngen_finemapping_df.groupBy(
"studyId", "region", "credibleSetIndex"
)
.agg(
f.collect_list(
f.struct(
f.col("variantId").cast("string").alias("variantId"),
f.col("posteriorProbability")
.cast("double")
.alias("posteriorProbability"),
f.col("logBF").cast("double").alias("logBF"),
f.col("pValueMantissa").cast("float").alias("pValueMantissa"),
f.col("pValueExponent").cast("integer").alias("pValueExponent"),
f.col("beta").cast("double").alias("beta"),
f.col("standardError").cast("double").alias("standardError"),
)
).alias("locus"),
)
.select(
"studyId",
"region",
"credibleSetIndex",
"locus",
)
.join(
toploci_df,
on=["studyId", "region", "credibleSetIndex"],
how="inner",
)
.withColumns(
{
"locusStart": f.split(f.split("region", ":")[1], "-")[0].cast(
"int"
),
"locusEnd": f.split(f.split("region", ":")[1], "-")[1].cast("int"),
}
)
).withColumn(
"studyLocusId",
StudyLocus.assign_study_locus_id(
["studyId", "variantId", "finemappingMethod"]
),
)
return StudyLocus(
_df=processed_finngen_finemapping_df,
_schema=StudyLocus.get_schema(),
).annotate_credible_sets()
|
