tcga-bulk-data-preprocessing-with-omicverse

TCGA bulk RNA-seq preprocessing with pyTCGA: GDC sample sheets, expression archives, clinical metadata, Kaplan-Meier survival analysis, and annotated AnnData export.

仓库: omicverse/omicclaw
最近来源活动: 2026年3月12日 11:02
检测到的 SKILL.md 语言: 英语
星标: 33
分支: 5

安装方式

默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。

检查来源文件

决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。

返回搜索返回安装

name	tcga-bulk-data-preprocessing-with-omicverse
title	TCGA bulk data preprocessing with omicverse
description	TCGA bulk RNA-seq preprocessing with pyTCGA: GDC sample sheets, expression archives, clinical metadata, Kaplan-Meier survival analysis, and annotated AnnData export.

TCGA Bulk Data Preprocessing with OmicVerse

Overview

Use this skill for loading TCGA data from GDC downloads, building normalised expression matrices, attaching clinical metadata, and running survival analyses through ov.bulk.pyTCGA.

Instructions

1. Gather required downloads

Confirm the user has three items from the GDC Data Portal:

gdc_sample_sheet.<date>.tsv — the sample sheet export
Decompressed gdc_download_xxxxx/ directory with expression archives
clinical.cart.<date>/ directory with clinical XML/JSON files

2. Initialise the TCGA helper

import omicverse as ov
import scanpy as sc
ov.plot_set()

aml_tcga = ov.bulk.pyTCGA(sample_sheet_path, download_dir, clinical_dir)
aml_tcga.adata_init()  # Builds AnnData with raw counts, FPKM, and TPM layers

3. Persist and reload

aml_tcga.adata.write_h5ad('data/ov_tcga_raw.h5ad', compression='gzip')

# To reload later:
new_tcga = ov.bulk.pyTCGA(sample_sheet_path, download_dir, clinical_dir)
new_tcga.adata_read()

aml_tcga.adata_meta_init()   # Gene ID → symbol mapping, patient info
aml_tcga.survial_init()      # NOTE: "survial" spelling — see Critical API Reference below

# Single gene
aml_tcga.survival_analysis('MYC', layer='deseq_normalize', plot=True)

# All genes (can take minutes for large gene sets)
aml_tcga.survial_analysis_all()  # NOTE: "survial" spelling

aml_tcga.adata.write_h5ad('data/ov_tcga_survival.h5ad', compression='gzip')

Method	Spelling	Purpose
`survial_init()`	survial (no 'v')	Initialize survival metadata columns
`survival_analysis(gene, layer, plot)`	survival (correct)	Single-gene Kaplan-Meier curve
`survial_analysis_all()`	survial (no 'v')	Sweep all genes for survival significance

# CORRECT — use the exact method names as documented
aml_tcga.survial_init()                    # "survial" — no 'v'
aml_tcga.survival_analysis('MYC', layer='deseq_normalize', plot=True)  # "survival" — correct
aml_tcga.survial_analysis_all()            # "survial" — no 'v'

# WRONG — these will raise AttributeError
# aml_tcga.survival_init()                 # AttributeError! Use survial_init()
# aml_tcga.survival_analysis_all()         # AttributeError! Use survial_analysis_all()

import os

# Before pyTCGA init: verify all paths exist
for name, path in [('sample_sheet', sample_sheet_path),
                    ('downloads', download_dir),
                    ('clinical', clinical_dir)]:
    if not os.path.exists(path):
        raise FileNotFoundError(f"TCGA {name} path not found: {path}")

# After adata_init(): verify expected layers were created
expected_layers = ['counts', 'fpkm', 'tpm']
for layer in expected_layers:
    if layer not in aml_tcga.adata.layers:
        print(f"WARNING: Missing layer '{layer}' — check if TCGA archives are fully extracted")

# Before survival analysis: verify metadata is initialized
if 'survial_init' not in dir(aml_tcga) or aml_tcga.adata.obs.shape[1] < 5:
    print("WARNING: Run adata_meta_init() and survial_init() before survival analysis")