聚类方法,聚类后选择打分最高那个分子,并对 karamadock 的结果求交集

scripts/cluster_granularity_scan.py

@@ -8,6 +8,40 @@ python scripts/cluster_granularity_scan.py \
     --smiles-col smiles \
     --radius 3 \
     --n-bits 1024
+
+Method          Params                    #Clusters  AvgSize    AvgIntraSim
+Butina          {'cutoff': 0.4}           8960       2.10       0.706     
+Butina          {'cutoff': 0.5}           6720       2.80       0.625     
+Butina          {'cutoff': 0.6}           4648       4.04       0.548     
+Butina          {'cutoff': 0.7}           2783       6.75       0.463     
+Butina          {'cutoff': 0.8}           958        19.61      0.333     
+Hierarchical    {'threshold': 0.3}        12235      1.54       0.814     
+Hierarchical    {'threshold': 0.4}        9603       1.96       0.739     
+Hierarchical    {'threshold': 0.5}        7300       2.57       0.664     
+DBSCAN          {'eps': 0.2}              2050       3.18       0.106     
+DBSCAN          {'eps': 0.3}              2275       4.61       0.113     
+DBSCAN          {'eps': 0.4}              2014       6.65       0.113     
+KMeans          {'k': 10}                 10         1878.70    0.204     
+KMeans          {'k': 20}                 20         939.35     0.200     
+KMeans          {'k': 50}                 50         375.74     0.233  
+| 列名              | 含义                                |
+| --------------- | --------------------------------- |
+| **#Clusters**   | 聚类后得到的簇数量（独立 cluster 数）           |
+| **AvgSize**     | 每个簇平均包含的分子个数 = 样本总数 / 簇数          |
+| **AvgIntraSim** | 每个簇内部分子两两之间的平均相似度（越接近 1 代表簇内部更相似） |
+现在的数据：
+
+Butina 在 cutoff=0.4 时 AvgIntraSim=0.706（簇内结构还算比较接近，但簇数非常多）。
+
+Hierarchical 阈值 0.3 时 AvgIntraSim=0.814（更紧密，但簇数更多）。
+
+DBSCAN 和 KMeans 的簇内相似度都低，说明它们在 Tanimoto 上可能不太适合你这个任务。
+
+聚类：用 Butina cutoff ≈ 0.6–0.7 或 Hierarchical 阈值 ≈ 0.5–0.6（保持簇内差异可控，簇数不要太多）。
+
+选代表：每个簇取 1 个中心分子（簇内与其他成员平均相似度最高的那个）。
+
+如果仍想增强多样性，可以在代表集中再跑一次 MaxMin picking。
 """
 import sys, os
 from pathlib import Path

scripts/example_api_usage.py

@@ -11,20 +11,20 @@ print("Running analysis examples...")
 
 # Example 1: Basic usage
 print("\nExample 1: Basic usage")
-main_api(['qed_values_fgbar.csv', 'qed_values_trpe.csv'], ['fgbar', 'trpe'])
+main_api(['finally_data/qed_values_poses_fgbar_all.csv', 'finally_data/qed_values_poses_trpe_all.csv'], ['fgbar', 'trpe'])
 
 # Example 2: With custom reference scores
 print("\nExample 2: With custom reference scores")
-main_api(['qed_values_fgbar.csv', 'qed_values_trpe.csv'], ['fgbar', 'trpe'], 
+main_api(['finally_data/qed_values_poses_fgbar_all.csv', 'finally_data/qed_values_poses_trpe_all.csv'], ['fgbar', 'trpe'], 
          reference_scores={'fgbar': {'9NY': -5.268}, 'trpe': {'0GA': -6.531}})
 
 # Example 3: With specific conformation rank
 print("\nExample 3: With specific conformation rank")
-main_api(['qed_values_fgbar.csv', 'qed_values_trpe.csv'], ['fgbar', 'trpe'], rank=0)
+main_api(['finally_data/qed_values_poses_fgbar_all.csv', 'finally_data/qed_values_poses_trpe_all.csv'], ['fgbar', 'trpe'], rank=0)
 
 # Example 4: With both custom reference scores and specific conformation rank
 print("\nExample 4: With both custom reference scores and specific conformation rank")
-main_api(['qed_values_fgbar.csv', 'qed_values_trpe.csv'], ['fgbar', 'trpe'], 
+main_api(['finally_data/qed_values_poses_fgbar_all.csv', 'finally_data/qed_values_poses_trpe_all.csv'], ['fgbar', 'trpe'], 
          reference_scores={'fgbar': {'9NY': -5.268}, 'trpe': {'0GA': -6.531}}, rank=0)
 
 print("\nAnalysis complete! Check the generated PNG files.")

scripts/extract_and_intersect.py

@@ -0,0 +1,59 @@
+import pandas as pd
+from pathlib import Path
+import os
+
+def process_cluster_file(cluster_file, score_file, output_file):
+    # 检查文件是否存在
+    if not os.path.exists(cluster_file):
+        raise FileNotFoundError(f"聚类文件不存在: {cluster_file}")
+    if not os.path.exists(score_file):
+        raise FileNotFoundError(f"评分文件不存在: {score_file}")
+    
+    # 读取聚类结果文件
+    cluster_df = pd.read_csv(cluster_file)
+    
+    # 提取filename列的stem属性
+    cluster_df['filename_stem'] = cluster_df['filename'].apply(
+        lambda x: Path(x).stem.split('_out')[0]
+    )
+    
+    # 读取score文件
+    score_df = pd.read_csv(score_file)
+    
+    # 获取两个文件的交集
+    intersection = pd.merge(
+        cluster_df, 
+        score_df, 
+        left_on='filename_stem', 
+        right_on='pdb_id',
+        how='inner'
+    )
+    
+    # 保存结果
+    intersection.to_csv(output_file, index=False)
+    
+    return len(intersection)
+
+if __name__ == "__main__":
+    # 使用绝对路径确保文件位置正确
+    base_dir = "/Users/lingyuzeng/Downloads/211.69.141.180/202508021824/vina"
+    
+    # 处理fgbar数据
+    fgbar_count = process_cluster_file(
+        f"{base_dir}/scripts/finally_data/cluster_best/fgbar_cluster_best_vina_butina_butina.csv",
+        f"{base_dir}/result/karamadock/FgBar1_score.csv",
+        f"{base_dir}/scripts/finally_data/cluster_best/fgbar_intersection.csv"
+    )
+    
+    # 处理trpe数据
+    trpe_count = process_cluster_file(
+        f"{base_dir}/scripts/finally_data/cluster_best/trpe_cluster_best_vina_butina_butina.csv",
+        f"{base_dir}/result/karamadock/TrpE_score.csv",
+        f"{base_dir}/scripts/finally_data/cluster_best/trpe_intersection.csv"
+    )
+    
+    print(f"fgbar交集数量: {fgbar_count}")
+    print(f"trpe交集数量: {trpe_count}")
+    
+    # 验证输出文件是否生成
+    print("脚本执行完成")

scripts/extract_top_molecules.py

@@ -0,0 +1,64 @@
+import pandas as pd
+import os
+import ast
+import argparse
+
+def parse_vina_scores(vina_scores_str):
+    """解析vina_scores字符串为浮点数列表"""
+    try:
+        scores = ast.literal_eval(vina_scores_str)
+        if isinstance(scores, list) and len(scores) > 0:
+            return scores[0]  # 取第一个值作为vina_score
+        return None
+    except:
+        return None
+
+def extract_top_molecules(file_path, output_dir, dataset_name):
+    """从CSV文件中提取karma_score_aligned和vina_score前1000的分子"""
+    # 读取数据
+    df = pd.read_csv(file_path)
+    
+    # 解析vina_scores列
+    df['vina_score'] = df['vina_scores'].apply(parse_vina_scores)
+    
+    # 按karma_score_aligned排序并提取前1000
+    df_karma_top = df.sort_values('karma_score_aligned', ascending=False).head(1000)
+    
+    # 按vina_score排序并提取前1000
+    df_vina_top = df.sort_values('vina_score', ascending=False).head(1000)
+    
+    # 保存结果
+    karma_output_file = os.path.join(output_dir, f"{dataset_name}_karma_score_aligned_top1000.csv")
+    vina_output_file = os.path.join(output_dir, f"{dataset_name}_vina_score_top1000.csv")
+    
+    df_karma_top.to_csv(karma_output_file, index=False)
+    df_vina_top.to_csv(vina_output_file, index=False)
+    
+    print(f"{dataset_name} - karma_score_aligned前1000分子保存到: {karma_output_file}")
+    print(f"{dataset_name} - vina_score前1000分子保存到: {vina_output_file}")
+    print(f"{dataset_name} - karma_score_aligned前1000分子数量: {len(df_karma_top)}")
+    print(f"{dataset_name} - vina_score前1000分子数量: {len(df_vina_top)}")
+    
+    return df_karma_top, df_vina_top
+
+def main():
+    parser = argparse.ArgumentParser(description='从CSV文件中提取karma_score_aligned和vina_score前1000的分子')
+    parser.add_argument('--input', nargs='+', required=True, 
+                        help='输入CSV文件路径列表')
+    parser.add_argument('--dataset-names', nargs='+', required=True,
+                        help='数据集名称列表，与输入文件一一对应')
+    parser.add_argument('--output', required=True,
+                        help='输出目录')
+    
+    args = parser.parse_args()
+    
+    # 确保输出目录存在
+    os.makedirs(args.output, exist_ok=True)
+    
+    # 处理每个文件
+    for file_path, dataset_name in zip(args.input, args.dataset_names):
+        print(f"Processing {dataset_name}...")
+        extract_top_molecules(file_path, args.output, dataset_name)
+
+if __name__ == "__main__":
+    main()