FunVIP 0.3.24.3__tar.gz → 0.3.24.5__tar.gz

{funvip-0.3.24.3 → funvip-0.3.24.5}/FunVIP.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: FunVIP
-Version: 0.3.24.3
+Version: 0.3.24.5
 Summary: Fungal Validation & Identification Pipeline
 Author-email: Changwan Seo <wan101010@snu.ac.kr>
 License: GPL-3.0

{funvip-0.3.24.3 → funvip-0.3.24.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: FunVIP
-Version: 0.3.24.3
+Version: 0.3.24.5
 Summary: Fungal Validation & Identification Pipeline
 Author-email: Changwan Seo <wan101010@snu.ac.kr>
 License: GPL-3.0

{funvip-0.3.24.3 → funvip-0.3.24.5}/funvip/src/cluster.py

@@ -155,11 +155,12 @@ def cluster(FI, V_list_group, V_cSR, path, opt):
             )
         elif group_count >= 2:
             logging.warning(
-                f"Query seq in {FI.id} has multiple matches to groups: {list(unique_groups)}"
+                f"Query seq in {FI.id} has multiple matches to groups within {opt.cluster.cutoff} cutoff: {list(unique_groups)}"
             )
-            FI.adjusted_group = next(
-                iter(unique_groups)
-            )  # Pick one (deterministic for testing)
+
+            # Among the multiple matches, get the best group
+            FI.adjusted_group = cutoff_df["subject_group"].iloc[0]
+
         else:
             logging.error("DEVELOPMENTAL ERROR IN GROUP ASSIGN")
             raise Exception

{funvip-0.3.24.3 → funvip-0.3.24.5}/funvip/src/concatenate.py

@@ -10,6 +10,7 @@ from funvip.src import search, hasher
 from scipy.optimize import minimize
 import numpy as np
 import shutil
+from time import time
 
 
 # Combine trimmed alignment of each gene to make concatenated matrix, and generate partition file
@@ -137,6 +138,8 @@ def combine_alignment(V, opt, path):
 def concatenate_df(V, path, opt):
     logging.info("Concatenating search results")
 
+    time_start = time()
+
     gene_list = []
     df_list = []
     for gene in V.dict_gene_SR.keys():
@@ -145,28 +148,6 @@ def concatenate_df(V, path, opt):
             gene_list.append(gene)
 
             df = deepcopy(V.dict_gene_SR[gene].set_index(["qseqid", "sseqid"]))
-
-            """
-            df = deepcopy(
-                V.dict_gene_SR[gene]
-                .set_index(["qseqid", "sseqid"])
-                .drop(
-                    columns=[
-                        "pident",
-                        "length",
-                        "mismatch",
-                        "gaps",
-                        "qstart",
-                        "qend",
-                        "sstart",
-                        "send",
-                        "evalue",
-                        "bitscore",
-                        "subject_group",
-                    ]
-                )
-            )
-            """
             df_list.append(df)
 
     if len(df_list) <= 0:
@@ -209,26 +190,22 @@ def concatenate_df(V, path, opt):
         )
 
         # Column name managing on subject_group
-        def same_merge(x, list_col):
-            values = x[list_col].dropna()
-            if values.empty:
-                raise Exception
-            return values.iloc[0]
 
-        df_multigene_regression_ori[
-            "subject_group"
-        ] = df_multigene_regression_ori.apply(
-            lambda x: same_merge(x, [f"{gene}_subject_group" for gene in gene_list]),
-            axis=1,
+        subject_group_cols = [f"{gene}_subject_group" for gene in gene_list]
+        df_multigene_regression_ori["subject_group"] = (
+            df_multigene_regression_ori[subject_group_cols].bfill(axis=1).iloc[:, 0]
         )
 
         # For regression, leave anchor points with all genes existing
         df_multigene_regression = df_multigene_regression_ori
 
-        for gene in gene_list:
-            df_multigene_regression = df_multigene_regression[
-                df_multigene_regression[f"{gene}_bitscore"].notna()
-            ]
+        mask = (
+            df_multigene_regression_ori[[f"{gene}_bitscore" for gene in gene_list]]
+            .notna()
+            .all(axis=1)
+        )
+
+        df_multigene_regression = df_multigene_regression_ori[mask]
 
         # Perform regression
         # Get regression line
@@ -248,23 +225,18 @@ def concatenate_df(V, path, opt):
             D = (P - (P.dot.u) * u).length
             """
             # line origin other than (0,0,0,..)
+            # do not use line -= l0 because casting rule error might happen
             if l0 is not None:
                 line = line - l0
             # points origin other than (0,0,0,..)
+            # do not use pts -= l0 because casting rule error might happen
             if p0 is not None:
                 pts = pts - p0
 
-            # dot product
-            dp = np.dot(pts, line)
-            # dot product value divided by normalized vector of line
-            # np.linalg.norm(line) : size of the line vector
-            # pp should be orthographic projected length of the dot
-            pp = dp / np.linalg.norm(line)
-            # norm value of point
-            # length from p0 to point
-            pn = np.linalg.norm(pts, axis=1)
-
-            return np.sqrt(np.clip(pn**2 - pp**2, a_min=1e-10, a_max=None))
+            norm_line = np.linalg.norm(line)
+            projected_lengths = np.dot(pts, line) / norm_line
+            distances = np.linalg.norm(pts, axis=1) ** 2 - projected_lengths**2
+            return np.sqrt(np.maximum(distances, 1e-10))
 
         # Optimization function
         def optimize_regression_line(points):
@@ -295,28 +267,16 @@ def concatenate_df(V, path, opt):
                 K_optimized,
             )
 
-        # Reset df before filling it
-        def calculate_prediction(row, gene_list, coeff, grad):
-            # Calculate linear_constant of the strain
-            linear_constant = []
+        def vectorized_prediction(df, gene_list, coeff, grad):
             for k, gene in enumerate(gene_list):
-                if not np.isnan(row[f"{gene}_bitscore"]):
-                    #  (coeff - value) / gradient = linear constant
-                    linear_constant.append(
-                        (coeff[k] - row[f"{gene}_bitscore"]) / grad[k]
-                    )
+                linear_constant = (coeff[k] - df[f"{gene}_bitscore"]) / grad[k]
+                mean_linear_constant = linear_constant.mean()
 
-            # Predict unknown BLAST value
-            for k, gene in enumerate(gene_list):
-                if np.isnan(row[f"{gene}_bitscore"]):
-                    #  prediction value  = coeff - linear constant * gradient
-                    prediction = coeff[k] - np.mean(linear_constant) * grad[k]
-                    row[f"{gene}_bitscore"] = prediction
-            return row
-
-        def apply_prediction(row, gene_list, coeff, grad):
-            row = calculate_prediction(row, gene_list, coeff, grad)
-            return row
+                # Fill missing bitscores
+                df[f"{gene}_bitscore"].fillna(
+                    coeff[k] - mean_linear_constant * grad[k], inplace=True
+                )
+            return df
 
         # Change to numpy for faster cazlculation
         np_bitscore = df_multigene_regression[
@@ -341,8 +301,9 @@ def concatenate_df(V, path, opt):
         # fill empty blast results for each gene with regression
         # This might be accelerated by using "apply", but coded manually initially because of logical complexity
         df_multigene_regression = df_multigene_regression_ori.copy()
-        df_multigene_regression = df_multigene_regression.apply(
-            apply_prediction, args=(gene_list, coeff, grad), axis=1
+
+        df_multigene_regression = vectorized_prediction(
+            df_multigene_regression, gene_list, coeff, grad
         )
 
         # Also update each gene bitscore matrix

{funvip-0.3.24.3 → funvip-0.3.24.5}/funvip/src/search.py

@@ -359,8 +359,8 @@ def search_df(V, path, opt):
                 )
 
                 # Cutoff by outgroupcutoff
-                print(vars(opt))
-                print(opt.cluster.outgroupoffset)
+                # print(vars(opt))
+                # print(opt.cluster.outgroupoffset)
                 df_search = df_search[
                     df_search["bitscore"] > opt.cluster.outgroupoffset
                 ]

{funvip-0.3.24.3 → funvip-0.3.24.5}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "FunVIP"
-version = "0.3.24.3"
+version = "0.3.24.5"
 description = "Fungal Validation & Identification Pipeline"
 authors = [{name = "Changwan Seo", email = "wan101010@snu.ac.kr"}]
 urls = { "Homepage" = "https://github.com/Changwanseo/FunVIP" }