Create cache folder
#3
by
pbordesinstadeep
- opened
multi_omics_transcript_expression.py
CHANGED
@@ -125,17 +125,6 @@ LABELS_V2 = [
|
|
125 |
"Whole Blood",
|
126 |
]
|
127 |
|
128 |
-
# Add after LABELS_V2 definition
|
129 |
-
LABELS_LIGHT = [
|
130 |
-
"Adipose Tissue",
|
131 |
-
"Brain",
|
132 |
-
"Heart",
|
133 |
-
"Liver",
|
134 |
-
"Lung",
|
135 |
-
"Muscle",
|
136 |
-
"Pancreas",
|
137 |
-
"Skin",
|
138 |
-
]
|
139 |
|
140 |
class GenomicLRATaskHandler(ABC):
|
141 |
"""
|
@@ -213,7 +202,6 @@ class GenomicLRATaskHandler(ABC):
|
|
213 |
|
214 |
if not os.path.exists(file_complete_path):
|
215 |
if not os.path.exists(file_complete_path + ".gz"):
|
216 |
-
os.makedirs(os.path.dirname(file_complete_path), exist_ok=True)
|
217 |
with tqdm(
|
218 |
unit="B",
|
219 |
unit_scale=True,
|
@@ -243,28 +231,31 @@ class TranscriptExpressionHandler(GenomicLRATaskHandler):
|
|
243 |
sequence_length: int = DEFAULT_LENGTH,
|
244 |
filter_out_sequence_length: int = DEFAULT_FILTER_OUT_LENGTH,
|
245 |
expression_method: str = "read_counts_old",
|
246 |
-
light_version: bool = False,
|
247 |
**kwargs,
|
248 |
):
|
249 |
"""
|
250 |
-
Creates a new handler for the
|
251 |
Args:
|
252 |
sequence_length: Length of the sequence around the TSS_CAGE start site
|
253 |
-
|
|
|
|
|
|
|
|
|
|
|
254 |
"""
|
255 |
self.reference_genome = None
|
256 |
self.coordinate_csv_file = None
|
257 |
self.labels_csv_file = None
|
258 |
-
self.light_version = light_version
|
259 |
self.sequence_length = sequence_length
|
260 |
self.filter_out_sequence_length = filter_out_sequence_length
|
261 |
|
262 |
-
if
|
263 |
-
assert isinstance(
|
264 |
assert (
|
265 |
-
|
266 |
-
), f"{
|
267 |
-
assert isinstance(
|
268 |
|
269 |
def get_info(self, description: str) -> DatasetInfo:
|
270 |
"""
|
@@ -294,7 +285,9 @@ class TranscriptExpressionHandler(GenomicLRATaskHandler):
|
|
294 |
}
|
295 |
)
|
296 |
return datasets.DatasetInfo(
|
|
|
297 |
description=description,
|
|
|
298 |
features=features,
|
299 |
)
|
300 |
|
@@ -327,30 +320,31 @@ class TranscriptExpressionHandler(GenomicLRATaskHandler):
|
|
327 |
"""
|
328 |
df = pd.read_csv(self.df_csv_file)
|
329 |
df = df.loc[df["chr"] != "chrMT"]
|
330 |
-
|
331 |
-
# Use light version labels if specified
|
332 |
-
labels_name = LABELS_LIGHT if self.light_version else LABELS_V1
|
333 |
|
334 |
split_df = df.loc[df["split"] == split]
|
335 |
-
|
336 |
-
# For light version, take only a subset of the data
|
337 |
-
if self.light_version:
|
338 |
-
split_df = split_df.sample(n=min(1000, len(split_df)), random_state=42)
|
339 |
|
340 |
norm_values_df = pd.read_csv(self.normalization_values_csv_file)
|
341 |
-
|
342 |
-
|
343 |
-
|
344 |
-
|
345 |
-
|
346 |
-
|
347 |
-
|
348 |
-
|
349 |
-
|
350 |
-
|
351 |
-
|
352 |
-
|
353 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
354 |
|
355 |
key = 0
|
356 |
for idx, coordinates_row in split_df.iterrows():
|
@@ -362,7 +356,7 @@ class TranscriptExpressionHandler(GenomicLRATaskHandler):
|
|
362 |
start = coordinates_row["start"] - 1 # -1 since vcf coords are 1-based
|
363 |
|
364 |
chromosome = coordinates_row["chr"]
|
365 |
-
labels_row = coordinates_row[
|
366 |
padded_sequence = pad_sequence(
|
367 |
chromosome=self.reference_genome[chromosome],
|
368 |
start=start,
|
@@ -508,4 +502,4 @@ def pad_sequence(
|
|
508 |
|
509 |
if negative_strand:
|
510 |
return chromosome[start:end].reverse.complement.seq
|
511 |
-
return chromosome[start:end].seq
|
|
|
125 |
"Whole Blood",
|
126 |
]
|
127 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
128 |
|
129 |
class GenomicLRATaskHandler(ABC):
|
130 |
"""
|
|
|
202 |
|
203 |
if not os.path.exists(file_complete_path):
|
204 |
if not os.path.exists(file_complete_path + ".gz"):
|
|
|
205 |
with tqdm(
|
206 |
unit="B",
|
207 |
unit_scale=True,
|
|
|
231 |
sequence_length: int = DEFAULT_LENGTH,
|
232 |
filter_out_sequence_length: int = DEFAULT_FILTER_OUT_LENGTH,
|
233 |
expression_method: str = "read_counts_old",
|
|
|
234 |
**kwargs,
|
235 |
):
|
236 |
"""
|
237 |
+
Creates a new handler for the Transcrpt Expression Prediction Task.
|
238 |
Args:
|
239 |
sequence_length: Length of the sequence around the TSS_CAGE start site
|
240 |
+
Instance Vars:
|
241 |
+
reference_genome: The Fasta extracted reference genome.
|
242 |
+
coordinate_csv_file: The csv file that stores the coordinates and filename of the target
|
243 |
+
labels_csv_file: The csv file that stores the labels with one sample per row.
|
244 |
+
sequence_length: Sequence length for this handler.
|
245 |
+
counts.
|
246 |
"""
|
247 |
self.reference_genome = None
|
248 |
self.coordinate_csv_file = None
|
249 |
self.labels_csv_file = None
|
|
|
250 |
self.sequence_length = sequence_length
|
251 |
self.filter_out_sequence_length = filter_out_sequence_length
|
252 |
|
253 |
+
if filter_out_sequence_length is not None:
|
254 |
+
assert isinstance(filter_out_sequence_length, int)
|
255 |
assert (
|
256 |
+
sequence_length <= filter_out_sequence_length
|
257 |
+
), f"{sequence_length=} > {filter_out_sequence_length=}"
|
258 |
+
assert isinstance(sequence_length, int)
|
259 |
|
260 |
def get_info(self, description: str) -> DatasetInfo:
|
261 |
"""
|
|
|
285 |
}
|
286 |
)
|
287 |
return datasets.DatasetInfo(
|
288 |
+
# This is the description that will appear on the datasets page.
|
289 |
description=description,
|
290 |
+
# This defines the different columns of the dataset and their types
|
291 |
features=features,
|
292 |
)
|
293 |
|
|
|
320 |
"""
|
321 |
df = pd.read_csv(self.df_csv_file)
|
322 |
df = df.loc[df["chr"] != "chrMT"]
|
323 |
+
labels_name = LABELS_V1
|
|
|
|
|
324 |
|
325 |
split_df = df.loc[df["split"] == split]
|
|
|
|
|
|
|
|
|
326 |
|
327 |
norm_values_df = pd.read_csv(self.normalization_values_csv_file)
|
328 |
+
m_t = (
|
329 |
+
norm_values_df[[f"m_t_{tissue}" for tissue in LABELS_V1]]
|
330 |
+
.to_numpy()
|
331 |
+
.reshape(-1)
|
332 |
+
)
|
333 |
+
sigma_t = (
|
334 |
+
norm_values_df[[f"sigma_t_{tissue}" for tissue in LABELS_V1]]
|
335 |
+
.to_numpy()
|
336 |
+
.reshape(-1)
|
337 |
+
)
|
338 |
+
m_g = (
|
339 |
+
norm_values_df[[f"m_g_{tissue}" for tissue in LABELS_V1]]
|
340 |
+
.to_numpy()
|
341 |
+
.reshape(-1)
|
342 |
+
)
|
343 |
+
sigma_g = (
|
344 |
+
norm_values_df[[f"sigma_g_{tissue}" for tissue in LABELS_V1]]
|
345 |
+
.to_numpy()
|
346 |
+
.reshape(-1)
|
347 |
+
)
|
348 |
|
349 |
key = 0
|
350 |
for idx, coordinates_row in split_df.iterrows():
|
|
|
356 |
start = coordinates_row["start"] - 1 # -1 since vcf coords are 1-based
|
357 |
|
358 |
chromosome = coordinates_row["chr"]
|
359 |
+
labels_row = coordinates_row[LABELS_V1]
|
360 |
padded_sequence = pad_sequence(
|
361 |
chromosome=self.reference_genome[chromosome],
|
362 |
start=start,
|
|
|
502 |
|
503 |
if negative_strand:
|
504 |
return chromosome[start:end].reverse.complement.seq
|
505 |
+
return chromosome[start:end].seq
|