sciv 0.0.88__py3-none-any.whl → 0.0.89__py3-none-any.whl

sciv/tool/_algorithm_.py

@@ -902,73 +902,97 @@ def overlap_sum(regions: AnnData, variants: dict, trait_info: DataFrame) -> AnnD
     """
 
     # Unique feature set
-    label_all = list(regions.var.index)
+    label_all = regions.var.index.tolist()
     # Peak number
     label_all_size: int = len(label_all)
 
-    # trait/disease information
-    trait_names: list = list(trait_info["id"])
+    # 预先把 peaks 的 index 做成 dict，O(1) 查找
+    label2idx = {lb: i for i, lb in enumerate(label_all)}
 
-    matrix = np.zeros((label_all_size, len(trait_names)))
+    trait_names = trait_info["id"].tolist()
+    n_trait = len(trait_names)
+    # 提前分配稀疏矩阵，按列填充，最后一次性转成 csc 再转 csr，省内存且快
+    row_indices, col_indices, data_vals = [], [], []
 
-    regions_df = regions.var.copy()
+    # 检查列存在性一次完成
+    required = {"chr", "start", "end"}
 
-    regions_columns: list = list(regions_df.columns)
-
-    if "chr" not in regions_columns or "start" not in regions_columns or "end" not in regions_columns:
+    if not required.issubset(regions.var.columns):
         ul.log(__name__).error(
-            f"The peaks information {regions_columns} in data `adata` must include three columns: `chr`, `start` and "
-            f"`end`. (It is recommended to use the `read_sc_atac` method.)"
+            f"The peaks information {regions.var.columns} in data `adata` must include three columns: `chr`, `start` "
+            f"and `end`. (It is recommended to use the `read_sc_atac` method.)"
         )
         raise ValueError(
-            f"The peaks information {regions_columns} in data `adata` must include three columns: `chr`, `start` and "
-            f"`end`. (It is recommended to use the `read_sc_atac` method.)"
+            f"The peaks information {regions.var.columns} in data `adata` must include three columns: `chr`, `start` "
+            f"and `end`. (It is recommended to use the `read_sc_atac` method.)"
         )
 
-    regions_df = regions_df.rename_axis("index")
-    regions_df = regions_df.reset_index()
-    # sort
-    regions_df = regions_df.sort_values(["chr", "start", "end"])[["index", "chr", "start", "end"]]
+    regions_df = (
+        regions.var
+        .reset_index()
+        .loc[:, ["index", "chr", "start", "end"]]
+        .sort_values(["chr", "start", "end"])
+    )
 
-    ul.log(__name__).info(f"Obtain peak-trait/disease matrix. (overlap variant information)")
-    for trait_name in tqdm(trait_names):
+    ul.log(__name__).info("Obtain peak-trait/disease matrix. (overlap variant information)")
 
+    # 外层循环按 trait 并行可再加速，这里先保持单循环
+    for col_idx, trait_name in enumerate(tqdm(trait_names)):
         variant: AnnData = variants[trait_name]
-        index: int = trait_names.index(trait_name)
+        overlap_df: DataFrame = _overlap_(regions_df, variant.obs)
 
-        # handle overlap data
-        overlap_info: DataFrame = _overlap_(regions_df, variant.obs)
-
-        if overlap_info.shape[0] == 0:
+        if overlap_df.empty:
             continue
 
-        overlap_info.rename({"index": "label"}, axis="columns", inplace=True)
-        overlap_info.reset_index(inplace=True)
-        overlap_info["region_id"] = (
-            overlap_info["chr"].astype(str)
-            + ":" + overlap_info["start"].astype(str) + "-" + overlap_info["end"].astype(str)
-        )
-
-        # get region
-        region_info = overlap_info.groupby("region_id", as_index=False)["label"].first()
-        region_info.index = region_info["label"].astype(str)
-        label: list = list(region_info["label"])
-
-        # Mutation information with repetitive features
-        label_size: int = len(label)
+        # 直接拿到 label->variant_id 的列表，省掉 groupby
+        overlap_df = overlap_df.rename(columns={"index": "label"})
+        # 把 label 映射到行号
+        overlap_df = overlap_df[overlap_df["label"].isin(label2idx)]
 
-        for j in range(label_size):
+        if overlap_df.empty:
+            continue
 
-            # Determine whether the features after overlap exist, In other words, whether there is overlap in this feature
-            if label[j] in label_all:
-                # get the index of label
-                label_index = label_all.index(label[j])
-                overlap_info_region = overlap_info[overlap_info["label"] == label[j]]
-                # sum value
-                overlap_variant = variant[list(overlap_info_region["variant_id"]), :]
-                matrix[label_index, index] = overlap_variant.X.sum(axis=0)
+        # 一次性求和：先按 label 分组，把 variant_id 收集成列表
+        label_var_ids = (
+            overlap_df
+            .groupby("label")["variant_id"]
+            .apply(list)
+            .reset_index()
+        )
 
-    overlap_adata = AnnData(to_sparse(matrix), var=trait_info, obs=regions.var)
+        # 遍历每个 label，一次性切片求和
+        for _, row in label_var_ids.iterrows():
+            label = row["label"]
+            row_idx = label2idx[label]
+            var_ids = row["variant_id"]
+            # 切片一次求和，避免逐行切片
+            matrix_sum = variant[var_ids, :].X.sum(axis=0)
+
+            if np.isscalar(matrix_sum):
+                matrix_sum = np.asarray(matrix_sum).reshape(1)
+
+            # 收集非零值
+            if matrix_sum.size == 1:
+                val = float(matrix_sum)
+                if val != 0:
+                    row_indices.append(row_idx)
+                    col_indices.append(col_idx)
+                    data_vals.append(val)
+            else:
+                for t_idx, v in enumerate(matrix_sum):
+                    if v != 0:
+                        row_indices.append(row_idx)
+                        col_indices.append(col_idx + t_idx)
+                        data_vals.append(float(v))
+
+    # 构建稀疏矩阵，再转 csr
+    overlap_sparse = sparse.csc_matrix(
+        (data_vals, (row_indices, col_indices)),
+        shape=(label_all_size, n_trait),
+        dtype=np.float32
+    ).tocsr()
+
+    overlap_adata = AnnData(overlap_sparse, var=trait_info, obs=regions.var)
     overlap_adata.uns["is_overlap"] = True
     return overlap_adata
 

sciv/tool/_random_walk_.py

@@ -201,6 +201,7 @@ def random_walk(
     if device == 'cpu' or (device == 'auto' and not availability):
         sample_count = seed_cell_weight.shape[1]
 
+        # 使用 joblib.Parallel 并指定 backend='threading' 保证顺序与输入一致
         results = Parallel(n_jobs=n_jobs)(
             delayed(_random_walk_cpu_)(seed_cell_weight[:, i], weight, gamma, epsilon, p)
             for i in tqdm(range(sample_count))
@@ -208,7 +209,19 @@ def random_walk(
 
         return np.column_stack(results)
     elif device == 'gpu' or (device == 'auto' and availability):
-        return _random_walk_gpu_(seed_cell_weight, weight, gamma, epsilon, p, device='gpu')
+
+        try:
+            return _random_walk_gpu_(seed_cell_weight, weight, gamma, epsilon, p, device='gpu')
+        except RuntimeError as e:
+            ul.log(__name__).warning(f"GPU failed to run, try to switch to CPU running.\n {e}")
+            sample_count = seed_cell_weight.shape[1]
+
+            results = Parallel(n_jobs=n_jobs)(
+                delayed(_random_walk_cpu_)(seed_cell_weight[:, i], weight, gamma, epsilon, p)
+                for i in tqdm(range(sample_count))
+            )
+
+            return np.column_stack(results)
     else:
         ul.log(__name__).error(f'The `device` ({device}) is not supported. Only supports "cpu", "gpu", and "auto" values.')
         raise ValueError(f'The `device` ({device}) is not supported. Only supports "cpu", "gpu", and "auto" values.')
@@ -677,7 +690,7 @@ class RandomWalk:
                 seed_cell_matrix_en[:, i] = seed_cell_en_value / (1 if seed_cell_en_value.sum() == 0 else seed_cell_en_value.sum())
 
         # Parallel processing of all traits and real-time display of progress
-        Parallel(n_jobs=-1, backend="threading")(
+        Parallel(n_jobs=self.n_jobs)(
             delayed(_process_single_trait)(i) for i in tqdm(self.trait_range, desc="Obtain progress of seed cells with weights")
         )
 
@@ -774,6 +787,7 @@ class RandomWalk:
 
         score = self._random_walk_core_(seed_cell_data, weight=weight)
 
+        ul.log(__name__).info("Normalize the results")
         cell_value = self.scale_norm(score)
 
         if _layer_label_ == "trs":
@@ -904,8 +918,8 @@ class RandomWalk:
         trs_score = to_dense(self.trs_adata.X if label == "run_en" else self.trs_adata.layers[_trs_layer_label_], is_array=True)
 
         # Initialize enriched container
-        trait_cell_enrichment = np.zeros(self.trs_adata.shape)
-        trait_cell_credible = np.zeros(self.trs_adata.shape)
+        trait_cell_enrichment = np.zeros(self.trs_adata.shape).astype(int)
+        trait_cell_credible = np.zeros(self.trs_adata.shape).astype(np.float32)
 
         ul.log(__name__).info(f"Calculate {len(self.trait_list)} traits/diseases for process `{label}`. (Enrichment-random walk)")
         # Random walk
@@ -916,8 +930,9 @@ class RandomWalk:
         )
 
         ul.log(__name__).info(f"Calculate {len(self.trait_list)} traits/diseases for process `{label}`. (Enrichment-score)")
-        for i in tqdm(self.trait_range):
 
+        # Process each trait in parallel
+        def _process_trait(i):
             # Random walk
             cell_value = cell_value_data[:, i]
 
@@ -940,7 +955,10 @@ class RandomWalk:
             trait_cell_enrichment[:, i][cell_value_credible > self.credible_threshold] = 1
             trait_cell_credible[:, i] = cell_value_credible
 
-        self.trs_adata.layers[_layer_label_] = to_sparse(trait_cell_enrichment.astype(int))
+        # Process each trait in parallel
+        Parallel(n_jobs=self.n_jobs)(delayed(_process_trait)(i) for i in tqdm(self.trait_range))
+
+        self.trs_adata.layers[_layer_label_] = to_sparse(trait_cell_enrichment)
 
         if not self.is_simple:
             self.trs_adata.layers[f"credible_{_layer_label_}"] = to_sparse(trait_cell_credible)

{sciv-0.0.88.dist-info → sciv-0.0.89.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sciv
-Version: 0.0.88
+Version: 0.0.89
 Summary: Unveiling the pivotal cell types involved in variant function regulation at a single-cell resolution
 Project-URL: github, https://github.com/YuZhengM/sciv
 Author-email: Zheng-Min Yu <yuzmbio@163.com>

{sciv-0.0.88.dist-info → sciv-0.0.89.dist-info}/RECORD

@@ -27,13 +27,13 @@ sciv/preprocessing/_scanpy_.py,sha256=flC10W5YPgJE5Ccxt5hP8qW3Q3cOi8DZ36Z1j9D3oN
 sciv/preprocessing/_scvi_.py,sha256=ZIDkQ_4deYmzSMiAbu5C3j_jMMl7hBTFLCBXHCNj3B4,10332
 sciv/preprocessing/_snapatac_.py,sha256=Dq8CHF7Psl3CQszaEokQYO56Oe2uzyWOy_cGlaOywfc,27798
 sciv/tool/__init__.py,sha256=WXzHkWt6RgBC3qqD-98nR5wQmt6oC850ox_VpMrapSU,2468
-sciv/tool/_algorithm_.py,sha256=fRzf7fjHF2O2beFEGBZX5qOF7nAKSdgHqHCUDZJ_3BE,49302
+sciv/tool/_algorithm_.py,sha256=-6YbvZEH95tvOicf2-bkvpmzdyOe5yAjelL7HjPaXF8,49952
 sciv/tool/_matrix_.py,sha256=O1EAhA9wxh06P_eOxEBesK7kO7IExKlhH6uJzGh1HBM,24322
-sciv/tool/_random_walk_.py,sha256=BKFjj5z1XI7Fzsr9Qsnzequ-oR5bMgK_WHf7KcpoSPU,47510
+sciv/tool/_random_walk_.py,sha256=27P1tjQtpjm44ZPE5CGEn3kb-PF-LSF2QsXw10jugSU,48304
 sciv/util/__init__.py,sha256=nOxZ8if27X7AUJ6hZwTwxOJwIBJb0obWlHjqCzjg_Gc,1964
 sciv/util/_constant_.py,sha256=w0wKQd8guLd1ZTW24_5aECrWsIWDiNQmEpLsWlHar1A,3000
 sciv/util/_core_.py,sha256=ZD2uSnEBHVu0i9TmXWzri_3bXZzYKnIZk818gW3zadE,14751
-sciv-0.0.88.dist-info/METADATA,sha256=V0-eiddGBywMSRVwQP8uPWs3Fq7OFpwpO4GYWkr_hqw,3465
-sciv-0.0.88.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-sciv-0.0.88.dist-info/licenses/LICENSE,sha256=4UvHVf3qCOZjHLs4LkYz8u96XRpXnZrpTKrkUQPs5_A,1075
-sciv-0.0.88.dist-info/RECORD,,
+sciv-0.0.89.dist-info/METADATA,sha256=-YZF4BLit8qmdZ_66DBLffciYokeFFWurTA0sEZi3IU,3465
+sciv-0.0.89.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+sciv-0.0.89.dist-info/licenses/LICENSE,sha256=4UvHVf3qCOZjHLs4LkYz8u96XRpXnZrpTKrkUQPs5_A,1075
+sciv-0.0.89.dist-info/RECORD,,