如何更简洁稳健地合并列名含部分匹配后缀的DataFrame？

需要合并两个Pandas DataFrame：tcia和clin。clin的列名末尾带有第三个短横线后的子串（例如TCGA-2K-A9WE-01，对应tcia中的TCGA-2K-A9WE），要求忽略该子串完成合并，最终保留tcia的列名格式。现有代码可实现需求，但希望获得更简洁稳健的实现方式。

现有代码：

clin_df = clin.copy()
clin_df.columns = clin_df.columns.str.rsplit('-', n=1).str.get(0)
df = pd.concat([clin_df, tcia], axis=0)
df.columns = clin.columns

输入数据 #

clin的结构 #

pd.DataFrame({'TCGA-2K-A9WE-01': {'admin.batch_number': '398.45.0',
  'age': '53',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': '207.0',
  'ethnicity': 'not hispanic or latino'},
 'TCGA-2Z-A9J1-01': {'admin.batch_number': '398.45.0',
  'age': '71',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': '2298.0',
  'ethnicity': 'not hispanic or latino'},
 'TCGA-2Z-A9J3-01': {'admin.batch_number': '398.45.0',
  'age': '67',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': nan,
  'ethnicity': 'not hispanic or latino'},
 'TCGA-2Z-A9J6-01': {'admin.batch_number': '398.45.0',
  'age': '60',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': '1731.0',
  'ethnicity': 'not hispanic or latino'},
 'TCGA-2Z-A9J7-01': {'admin.batch_number': '398.45.0',
  'age': '63',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': nan,
  'ethnicity': 'not hispanic or latino'}})

tcia的结构 #

pd.DataFrame({'TCGA-2K-A9WE': {'ips_ctla4_neg_pd1_neg': 8.0,
  'ips_ctla4_neg_pd1_pos': 7.0,
  'ips_ctla4_pos_pd1_neg': 7.0,
  'ips_ctla4_pos_pd1_pos': 6.0,
  'patient_uuid': '73292c19-d6a8-4bc4-97bc-ccce54f264f8'},
 'TCGA-2Z-A9J1': {'ips_ctla4_neg_pd1_neg': 9.0,
  'ips_ctla4_neg_pd1_pos': 8.0,
  'ips_ctla4_pos_pd1_neg': 9.0,
  'ips_ctla4_pos_pd1_pos': 7.0,
  'patient_uuid': '851a1157-e460-4794-8534-2eb6f0ae7468'},
 'TCGA-2Z-A9J3': {'ips_ctla4_neg_pd1_neg': 9.0,
  'ips_ctla4_neg_pd1_pos': 7.0,
  'ips_ctla4_pos_pd1_neg': 8.0,
  'ips_ctla4_pos_pd1_pos': 6.0,
  'patient_uuid': '5195c9ac-b649-49f8-8750-f9a4787e8e52'},
 'TCGA-2Z-A9J6': {'ips_ctla4_neg_pd1_neg': 9.0,
  'ips_ctla4_neg_pd1_pos': 7.0,
  'ips_ctla4_pos_pd1_neg': 8.0,
  'ips_ctla4_pos_pd1_pos': 7.0,
  'patient_uuid': '4a540448-f106-4b0e-9038-9f7ccefc785b'},
 'TCGA-2Z-A9J7': {'ips_ctla4_neg_pd1_neg': 7.0,
  'ips_ctla4_neg_pd1_pos': 5.0,
  'ips_ctla4_pos_pd1_neg': 6.0,
  'ips_ctla4_pos_pd1_pos': 5.0,
  'patient_uuid': 'd66c9261-6c0c-44b0-92fa-a43757f34cb2'}})

期望输出 #

pd.DataFrame({'TCGA-2K-A9WE': {'admin.batch_number': '398.45.0',
  'age': '53',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': '207.0',
  'ethnicity': 'not hispanic or latino',
  'ips_ctla4_neg_pd1_neg': 8.0,
  'ips_ctla4_neg_pd1_pos': 7.0,
  'ips_ctla4_pos_pd1_neg': 7.0,
  'ips_ctla4_pos_pd1_pos': 6.0,
  'patient_uuid': '73292c19-d6a8-4bc4-97bc-ccce54f264f8'},
 'TCGA-2Z-A9J1': {'admin.batch_number': '398.45.0',
  'age': '71',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': '2298.0',
  'ethnicity': 'not hispanic or latino',
  'ips_ctla4_neg_pd1_neg': 9.0,
  'ips_ctla4_neg_pd1_pos': 8.0,
  'ips_ctla4_pos_pd1_neg': 9.0,
  'ips_ctla4_pos_pd1_pos': 7.0,
  'patient_uuid': '851a1157-e460-4794-8534-2eb6f0ae7468'},
 'TCGA-2Z-A9J3': {'admin.batch_number': '398.45.0',
  'age': '67',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': nan,
  'ethnicity': 'not hispanic or latino',
  'ips_ctla4_neg_pd1_neg': 9.0,
  'ips_ctla4_neg_pd1_pos': 7.0,
  'ips_ctla4_pos_pd1_neg': 8.0,
  'ips_ctla4_pos_pd1_pos': 6.0,
  'patient_uuid': '5195c9ac-b649-49f8-8750-f9a4787e8e52'},
 'TCGA-2Z-A9J6': {'admin.batch_number': '398.45.0',
  'age': '60',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': '1731.0',
  'ethnicity': 'not hispanic or latino',
  'ips_ctla4_neg_pd1_neg': 9.0,
  'ips_ctla4_neg_pd1_pos': 7.0,
  'ips_ctla4_pos_pd1_neg': 8.0,
  'ips_ctla4_pos_pd1_pos': 7.0,
  'patient_uuid': '4a540448-f106-4b0e-9038-9f7ccefc785b'},
 'TCGA-2Z-A9J7': {'admin.batch_number': '398.45.0',
  'age': '63',
  'days_to_initial_pathologic_diagnosis': '0',
  'days_to_last_follow_up': nan,
  'ethnicity': 'not hispanic or latino',
  'ips_ctla4_neg_pd1_neg': 7.0,
  'ips_ctla4_neg_pd1_pos': 5.0,
  'ips_ctla4_pos_pd1_neg': 6.0,
  'ips_ctla4_pos_pd1_pos': 5.0,
  'patient_uuid': 'd66c9261-6c0c-44b0-92fa-a43757f34cb2'}})

直接通过重命名clin的列并与tcia横向合并的方式实现，逻辑更清晰，避免冗余操作：

# 重命名clin的列，去掉最后一段后缀
clin_renamed = clin.rename(columns=lambda col: col.rsplit('-', n=1)[0])
# 转置后按索引合并，再转回原结构
merged_df = clin_renamed.T.join(tcia.T).T

也可以用链式调用简化代码：

merged_df = clin.rename(columns=lambda col: col.rsplit('-', n=1)[0]).T.join(tcia.T).T

优化说明 #

1. 逻辑更直观：原代码用行拼接后强行修改列名，绕了弯路；直接通过join按处理后的列名对齐合并，符合数据合并的常规逻辑。
2. 减少冗余操作：省略了不必要的copy()（rename默认返回新对象），也避免了行拼接后可能出现的索引混乱问题。
3. 稳健性提升：使用lambda函数重命名列，比直接修改columns属性更灵活，能兼容列名格式不一致的边界情况。

内容的提问来源于stack exchange，提问作者Anon