Liu-Hy's picture
Add files using upload-large-folder tool
24fe0da verified
# Path Configuration
from tools.preprocess import *
# Processing context
trait = "Epilepsy"
cohort = "GSE64123"
# Input paths
in_trait_dir = "../DATA/GEO/Epilepsy"
in_cohort_dir = "../DATA/GEO/Epilepsy/GSE64123"
# Output paths
out_data_file = "./output/preprocess/1/Epilepsy/GSE64123.csv"
out_gene_data_file = "./output/preprocess/1/Epilepsy/gene_data/GSE64123.csv"
out_clinical_data_file = "./output/preprocess/1/Epilepsy/clinical_data/GSE64123.csv"
json_path = "./output/preprocess/1/Epilepsy/cohort_info.json"
# STEP1
from tools.preprocess import *
# 1. Identify the paths to the SOFT file and the matrix file
soft_file, matrix_file = geo_get_relevant_filepaths(in_cohort_dir)
# 2. Read the matrix file to obtain background information and sample characteristics data
background_prefixes = ['!Series_title', '!Series_summary', '!Series_overall_design']
clinical_prefixes = ['!Sample_geo_accession', '!Sample_characteristics_ch1']
background_info, clinical_data = get_background_and_clinical_data(matrix_file, background_prefixes, clinical_prefixes)
# 3. Obtain the sample characteristics dictionary from the clinical dataframe
sample_characteristics_dict = get_unique_values_by_row(clinical_data)
# 4. Explicitly print out all the background information and the sample characteristics dictionary
print("Background Information:")
print(background_info)
print("Sample Characteristics Dictionary:")
print(sample_characteristics_dict)
# 1. Gene Expression Data Availability
# Based on the background info and the sample dictionary,
# this dataset is most likely a gene expression study (not miRNA or methylation).
is_gene_available = True
# 2. Variable Availability and Data Type Conversion
# All rows in the sample characteristics dictionary pertain to time, drug exposure, or concentration.
# There is no row indicating a human trait of "Epilepsy," nor any human "age" or "gender."
trait_row = None
age_row = None
gender_row = None
is_trait_available = (trait_row is not None)
# Although we found no data for these variables, we still define stub conversion functions:
def convert_trait(value: str):
return None # No trait data available
def convert_age(value: str):
return None # No age data available
def convert_gender(value: str):
return None # No gender data available
# 3. Save Metadata (Initial Filtering)
# Because it's the initial filtering, use is_final=False.
validate_and_save_cohort_info(
is_final=False,
cohort=cohort,
info_path=json_path,
is_gene_available=is_gene_available,
is_trait_available=is_trait_available
)
# 4. Clinical Feature Extraction
# This step is skipped because 'trait_row' is None (no trait data).
# STEP3
# 1. Use the get_genetic_data function from the library to get the gene_data from the matrix_file previously defined.
gene_data = get_genetic_data(matrix_file)
# 2. Print the first 20 row IDs (gene or probe identifiers) for future observation.
print(gene_data.index[:20])
# These probe IDs (e.g., "10000_at") are not standard human gene symbols, thus mapping is required.
print("requires_gene_mapping = True")
# STEP5
import pandas as pd
import io
# 1. Extract the lines that do NOT start with '^', '!', or '#', but do NOT parse them into a DataFrame yet.
annotation_text, _ = filter_content_by_prefix(
source=soft_file,
prefixes_a=['^', '!', '#'],
unselect=True,
source_type='file',
return_df_a=False,
return_df_b=False
)
# 2. Manually parse the filtered text into a DataFrame, specifying engine="python" to avoid buffer overflow issues.
gene_annotation = pd.read_csv(
io.StringIO(annotation_text),
delimiter='\t',
on_bad_lines='skip',
engine='python'
)
print("Gene annotation preview:")
print(preview_df(gene_annotation))
# STEP: Gene Identifier Mapping
# 1. Determine which columns in gene_annotation match the data in gene_data.
# - The probe IDs in gene_data match the "ID" column in gene_annotation.
# - The gene symbols appear to be in the "Description" column.
# 2. Create the mapping dataframe using the library function get_gene_mapping.
mapping_df = get_gene_mapping(gene_annotation, prob_col="ID", gene_col="Description")
# 3. Convert the probe-level measurements in gene_data to gene-level expressions.
gene_data = apply_gene_mapping(gene_data, mapping_df)
import os
import pandas as pd
# STEP7
# 1) Normalize gene symbols and save
normalized_gene_data = normalize_gene_symbols_in_index(gene_data)
normalized_gene_data.to_csv(out_gene_data_file)
# Check whether we actually have a clinical CSV file (i.e., trait data) from Step 2
if os.path.exists(out_clinical_data_file):
# 2) Link the clinical and gene expression data
selected_clinical_df = pd.read_csv(out_clinical_data_file, index_col=0)
linked_data = geo_link_clinical_genetic_data(selected_clinical_df, normalized_gene_data)
# 3) Handle missing values
final_data = handle_missing_values(linked_data, trait_col=trait)
# 4) Evaluate bias in the trait
trait_biased, final_data = judge_and_remove_biased_features(final_data, trait)
# 5) Final validation (trait is available)
is_usable = validate_and_save_cohort_info(
is_final=True,
cohort=cohort,
info_path=json_path,
is_gene_available=True,
is_trait_available=True,
is_biased=trait_biased,
df=final_data,
note="Trait data successfully extracted in Step 2."
)
# 6) If the dataset is usable, save
if is_usable:
final_data.to_csv(out_data_file)
else:
# If the clinical file does not exist, the trait is unavailable
# Perform final validation indicating that we lack trait data
empty_df = pd.DataFrame()
validate_and_save_cohort_info(
is_final=True,
cohort=cohort,
info_path=json_path,
is_gene_available=True,
is_trait_available=False,
is_biased=True, # Arbitrary non-None to skip usage
df=empty_df,
note="No trait data was found; linking and final dataset output are skipped."
)