如何基于Pandas DataFrame绘制散点图将药物按靶点聚类为7组

现有如下包含target、drugs两列的Pandas DataFrame：

target                        drugs
0   ACE2 gene                [angiotensin II,rosiglitazone,irbesartan,valsar]
1   Elastases                [heparin,prednisolone,montelukast,formoterol]
2   MAPK14 protein           [oxaprozin,nilotinib,imatinib,tocilizumab]
3   TMPRSS2 gene             [enzalutamide,camostat]
4   Toll-like receptors      [rituximab,atorvastatin,artesunate,tazarotene]
5   Ubiquitin                [sunitinib,lapatinib,atorvastatin,edaravone]
6   ezrin                    [erlotinib,crizotinib,sorafenib,everolimus]

需求为绘制散点图，将所有药物按所属靶点聚类，最终得到7个对应不同靶点的聚类分组。

生成该DataFrame的参考代码如下：

import pandas as pd
    
data = {'target': ['ACE2 gene', 'Elastases', 'MAPK14 protein, human', 'TMPRSS2 gene', 'Toll-like receptors', 'Ubiquitin' , 'ezrin'],'drugs': [['angiotensin II','rosiglitazone','irbesartan'], ['heparin','prednisolone','montelukast','formoterol'] , ['oxaprozin','nilotinib','imatinib','tocilizumab'] , ['enzalutamide','camostat'] , ['rituximab','atorvastatin','artesunate','tazarotene'] , ['sunitinib','lapatinib','atorvastatin','edaravone'], ['erlotinib','crizotinib','sorafenib','everolimus']]
}
    
df = pd.DataFrame(data)

以下是可直接运行的实现代码：

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np

# 1. 生成原始数据
data = {'target': ['ACE2 gene', 'Elastases', 'MAPK14 protein, human', 'TMPRSS2 gene', 'Toll-like receptors', 'Ubiquitin' , 'ezrin'],'drugs': [['angiotensin II','rosiglitazone','irbesartan'], ['heparin','prednisolone','montelukast','formoterol'] , ['oxaprozin','nilotinib','imatinib','tocilizumab'] , ['enzalutamide','camostat'] , ['rituximab','atorvastatin','artesunate','tazarotene'] , ['sunitinib','lapatinib','atorvastatin','edaravone'], ['erlotinib','crizotinib','sorafenib','everolimus']]
}
df = pd.DataFrame(data)

# 2. 展开药物列表，每行对应一个药物+所属靶点
df_explode = df.explode('drugs', ignore_index=True)

# 3. 给每个靶点分配中心坐标，保证不同靶点中心距离足够远
targets = df_explode['target'].unique()
n_targets = len(targets)
# 生成7个离散的中心坐标
center_coords = np.array([[i%3, i//3] for i in range(n_targets)])
target_to_center = dict(zip(targets, center_coords))

# 4. 给每个药物在对应靶点中心附近加小范围随机噪声，形成聚类效果
np.random.seed(42) # 固定随机种子保证结果可复现
df_explode['x'] = df_explode['target'].apply(lambda t: target_to_center[t][0] + np.random.normal(0, 0.1))
df_explode['y'] = df_explode['target'].apply(lambda t: target_to_center[t][1] + np.random.normal(0, 0.1))

# 5. 绘制散点图
plt.figure(figsize=(10,7))
for target in targets:
    subset = df_explode[df_explode['target'] == target]
    plt.scatter(subset['x'], subset['y'], label=target, s=100, alpha=0.8)
# 可选：给每个点标注药物名称
for _, row in df_explode.iterrows():
    plt.text(row['x']+0.02, row['y']+0.02, row['drugs'], fontsize=8)

plt.legend(bbox_to_anchor=(1.05, 1), loc='upper left')
plt.title('药物按靶点聚类散点图')
plt.axis('off') # 隐藏坐标轴更美观
plt.tight_layout()
plt.show()

效果说明 #

• 同靶点的药物会自动聚集在同一个区域，一共形成7个明确的聚类分组
• 不同靶点用不同颜色区分，可通过图例对应靶点名称
• 每个散点可选择标注对应药物名称，方便查看对应关系

内容的提问来源于stack exchange，提问作者Electra