workbench 0.8.172__tar.gz → 0.8.174__tar.gz

{workbench-0.8.172 → workbench-0.8.174}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: workbench
-Version: 0.8.172
+Version: 0.8.174
 Summary: Workbench: A Dashboard and Python API for creating and deploying AWS SageMaker Model Pipelines
 Author-email: SuperCowPowers LLC <support@supercowpowers.com>
 License-Expression: MIT

{workbench-0.8.172 → workbench-0.8.174}/aws_setup/workbench_core/workbench_core/workbench_core_stack.py

@@ -820,6 +820,17 @@ class WorkbenchCoreStack(Stack):
             resources=[f"arn:aws:sagemaker:{self.region}:{self.account}:monitoring-schedule/*"],
         )
 
+    def endpoint_monitoring_processing(self) -> iam.PolicyStatement:
+        """Processing jobs for Model Monitor baselines and executions."""
+        return iam.PolicyStatement(
+            actions=[
+                "sagemaker:CreateProcessingJob",
+                "sagemaker:DescribeProcessingJob",
+                "sagemaker:StopProcessingJob",
+            ],
+            resources=[f"arn:aws:sagemaker:{self.region}:{self.account}:processing-job/*"],
+        )
+
     def endpoint_data_quality(self) -> iam.PolicyStatement:
         """Create a policy statement for managing SageMaker endpoint data quality jobs.
 
@@ -881,54 +892,6 @@ class WorkbenchCoreStack(Stack):
             resources=[sns_resources],
         )
 
-    @staticmethod
-    def pipeline_list() -> iam.PolicyStatement:
-        """Create a policy statement for listing SageMaker pipelines.
-
-        Returns:
-            iam.PolicyStatement: The policy statement for listing SageMaker pipelines.
-        """
-        return iam.PolicyStatement(
-            actions=[
-                "sagemaker:ListPipelines",
-            ],
-            resources=["*"],  # Broad permission necessary for listing operations
-        )
-
-    def pipeline_full(self) -> iam.PolicyStatement:
-        """Create a policy statement for inspecting and running SageMaker Pipelines.
-
-        Returns:
-            iam.PolicyStatement: The policy statement for inspecting and running SageMaker Pipelines.
-        """
-        pipeline_resources = f"arn:aws:sagemaker:{self.region}:{self.account}:pipeline/*"
-        execution_resources = f"arn:aws:sagemaker:{self.region}:{self.account}:pipeline-execution/*"
-        processing_resources = f"arn:aws:sagemaker:{self.region}:{self.account}:processing-job/*"
-
-        return iam.PolicyStatement(
-            actions=[
-                "sagemaker:DescribePipeline",
-                "sagemaker:ListPipelineExecutions",
-                "sagemaker:DescribePipelineExecution",
-                "sagemaker:ListPipelineExecutionSteps",
-                "sagemaker:StartPipelineExecution",
-                # Actions for jobs
-                "sagemaker:CreateProcessingJob",
-                "sagemaker:DescribeProcessingJob",
-                "sagemaker:ListProcessingJobs",
-                "sagemaker:StopProcessingJob",
-                # Tagging
-                "sagemaker:ListTags",
-                "sagemaker:AddTags",
-                "sagemaker:DeleteTags",
-            ],
-            resources=[
-                pipeline_resources,
-                execution_resources,
-                processing_resources,
-            ],
-        )
-
     @staticmethod
     def ecr_policy_statement() -> iam.PolicyStatement:
         """Create a policy statement for pulling ECR Images.
@@ -1284,6 +1247,7 @@ class WorkbenchCoreStack(Stack):
             self.endpoint_read(),
             self.endpoint_monitoring_discovery(),
             self.endpoint_monitoring_schedules(),
+            self.endpoint_monitoring_processing(),
             self.cloudwatch_logs(),
             self.cloudwatch_metrics(),
             self.parameter_store_discover(),
@@ -1306,6 +1270,7 @@ class WorkbenchCoreStack(Stack):
             self.endpoint_data_quality(),
             self.endpoint_monitoring_discovery(),
             self.endpoint_monitoring_schedules(),
+            self.endpoint_monitoring_processing(),
             self.cloudwatch_logs(),
             self.cloudwatch_metrics(),
             self.parameter_store_discover(),
@@ -1318,19 +1283,6 @@ class WorkbenchCoreStack(Stack):
             managed_policy_name="WorkbenchEndpointPolicy",
         )
 
-    def sagemaker_full_policy_not_used(self) -> iam.ManagedPolicy:
-        """Create a managed policy for the Workbench Pipelines (FULL)"""
-        policy_statements = [
-            self.pipeline_list(),
-            self.pipeline_full(),
-        ]
-        return iam.ManagedPolicy(
-            self,
-            id="WorkbenchSagemakerPipelineFullPolicy",
-            statements=policy_statements,
-            managed_policy_name="WorkbenchSagemakerPipelineFullPolicy",
-        )
-
     def workbench_dataframe_store_read_policy(self) -> iam.ManagedPolicy:
         """Create a managed policy for the Workbench DataFrame Store (READ-ONLY)"""
         policy_statements = [

{workbench-0.8.172 → workbench-0.8.174}/docs/api_classes/monitor.md

@@ -18,7 +18,7 @@ endpoint_name = "abalone-regression-end-rt"
 mon = Monitor(endpoint_name)
 
 # Add data capture to the endpoint
-mon.add_data_capture(capture_percentage=100)
+mon.enable_data_capture(capture_percentage=100)
 
 # Create a baseline for monitoring
 mon.create_baseline()
@@ -34,7 +34,7 @@ from workbench.api.monitor import Monitor
 from workbench.api.endpoint import Endpoint
 
 # Construct a Monitor Class in one of Two Ways
-mon = Endpoint("abalone-regression-end-rt").get_monitor()
+mon = Endpoint("abalone-regression-end-rt").monitor()
 mon = Monitor("abalone-regression-end-rt")
 
 # Check the summary and details of the monitoring class

{workbench-0.8.172 → workbench-0.8.174}/examples/chem_utils/tautomerize_smiles.py

@@ -1,7 +1,7 @@
 """Example for Tautomerizing SMILES strings"""
 
 import pandas as pd
-from workbench.utils.chem_utils import tautomerize_smiles
+from workbench.utils.chem_utils.mol_standardize import standardize
 
 pd.options.display.max_columns = None
 pd.options.display.width = 1200
@@ -32,6 +32,12 @@ test_data = [
 # Convert test data to a DataFrame
 df = pd.DataFrame(test_data)
 
-# Perform tautomerization
-result_df = tautomerize_smiles(df)
-print(result_df)
+# Perform tautomerization using standardize with tautomer canonicalization
+result_df = standardize(df)
+
+# Add comparison column
+result_df["matches_expected"] = result_df["smiles"] == result_df["expected"]
+
+# Display results
+print("Tautomerization Results:")
+print(result_df[["id", "orig_smiles", "smiles", "expected", "matches_expected"]])

workbench-0.8.174/examples/models/smiles_to_md_v1.py

@@ -0,0 +1,27 @@
+"""Create the SMILES to Standardize + Tautomer + Molecular Descriptors + Stereo Model and Endpoint"""
+
+# Workbench Imports
+from workbench.api import FeatureSet, ModelType, Model
+from workbench.utils.model_utils import get_custom_script_path
+
+# fs_name = "aqsol_features"
+fs_name = "solubility_featurized_class_0_fs"
+
+
+script_path = get_custom_script_path("chem_info", "molecular_descriptors.py")
+feature_set = FeatureSet(fs_name)
+tags = ["smiles", "molecular descriptors", "tautomerized", "stereo"]
+model = feature_set.to_model(
+    name="smiles-to-taut-md-stereo-v1",
+    model_type=ModelType.TRANSFORMER,
+    feature_list=["smiles"],
+    description="Smiles to Molecular Descriptors",
+    tags=tags,
+    custom_script=script_path,
+)
+
+# Create the endpoint for the model
+end = model.to_endpoint(tags=tags, mem_size=4096, max_concurrency=10)
+
+# Run inference on the endpoint
+end.auto_inference(capture=True)

{workbench-0.8.172 → workbench-0.8.174}/examples/monitor/monitor_usage.py

@@ -3,7 +3,7 @@ from workbench.api.monitor import Monitor
 from workbench.api.endpoint import Endpoint
 
 # Construct a Monitor Class in one of Two Ways
-mon = Endpoint("abalone-regression-end-rt").get_monitor()
+mon = Endpoint("abalone-regression-end-rt").monitor()
 mon = Monitor("abalone-regression-end-rt")
 
 # Check the summary and details of the monitoring class
@@ -19,8 +19,3 @@ print(constraints_df.head())
 
 statistics_df = mon.get_statistics()
 print(statistics_df.head())
-
-# Get the latest data capture (inputs and outputs)
-input_df, output_df = mon.get_captured_data()
-print(input_df.head())
-print(output_df.head())

{workbench-0.8.172 → workbench-0.8.174}/experiments/chem_info/bulk_sims.py

@@ -1,6 +1,6 @@
 # Workbench Imports
 from workbench.api.df_store import DFStore
-from workbench.utils.chem_utils import compute_morgan_fingerprints
+from workbench.utils.chem_utils.fingerprints import compute_morgan_fingerprints
 from rdkit.Chem import DataStructs
 from rdkit.DataStructs.cDataStructs import CreateFromBitString
 import pandas as pd

{workbench-0.8.172 → workbench-0.8.174}/experiments/chem_info/knn_sims.py

@@ -1,6 +1,6 @@
 # Workbench Imports
 from workbench.api.df_store import DFStore
-from workbench.utils.chem_utils import compute_morgan_fingerprints
+from workbench.utils.chem_utils.fingerprints import compute_morgan_fingerprints
 from workbench.algorithms.dataframe.fingerprint_proximity import FingerprintProximity
 
 # Grab tox21 dataset (only 10)

{workbench-0.8.172 → workbench-0.8.174}/scripts/tox21/test_tox_heuristic.py

@@ -1,6 +1,6 @@
 from rdkit import Chem
 from workbench.api.df_store import DFStore
-from workbench.utils.chem_utils import toxic_elements, toxic_groups
+from workbench.utils.chem_utils.toxicity import toxic_elements, toxic_groups
 
 # Grab the tox21 data
 tox21 = DFStore().get("/datasets/chem_info/tox21")

workbench-0.8.174/scripts/tox21/tox21_to_df_store.py

@@ -0,0 +1,186 @@
+import pandas as pd
+from rdkit import Chem
+from rdkit.Chem import PandasTools
+
+# Workbench Imports
+from workbench.api.df_store import DFStore
+
+# Import the new mol_tagging module
+from workbench.utils.chem_utils.mol_tagging import tag_molecules, filter_by_tags, get_tag_summary
+from workbench.utils.chem_utils.fingerprints import compute_morgan_fingerprints
+from workbench.utils.chem_utils.projections import project_fingerprints
+
+
+def prep_sdf_file(filepath: str) -> pd.DataFrame:
+    """
+    Prepare SDF file for analysis using the new tagging system.
+    """
+
+    # Load SDF file directly into a DataFrame
+    df = PandasTools.LoadSDF(filepath, smilesName="smiles", molColName="molecule", includeFingerprints=False)
+    print(f"Loaded {len(df)} compounds from {filepath}")
+    print(f"Columns: {df.columns.tolist()}")
+
+    # Standardize column names
+    df.rename(columns={"ID": "id"}, inplace=True)
+
+    # Convert Formula weight to numeric
+    df["FW"] = pd.to_numeric(df["FW"], errors="coerce")
+
+    # Full list of assay columns for toxicity detection
+    assay_cols = [
+        "NR-AR",  # Nuclear Receptor - Androgen Receptor
+        "SR-HSE",  # Stress Response - Heat Shock Element
+        "SR-ARE",  # Stress Response - Antioxidant Response Element
+        "NR-Aromatase",  # Nuclear Receptor - Aromatase
+        "NR-ER-LBD",  # Nuclear Receptor - Estrogen Receptor (Ligand Binding Domain)
+        "NR-AhR",  # Nuclear Receptor - Aryl Hydrocarbon Receptor
+        "SR-MMP",  # Stress Response - Mitochondrial Membrane Potential
+        "NR-ER",  # Nuclear Receptor - Estrogen Receptor
+        "NR-PPAR-gamma",  # Nuclear Receptor - Peroxisome Proliferator-Activated Receptor Gamma
+        "SR-p53",  # Stress Response - p53 Tumor Suppressor Protein
+        "SR-ATAD5",  # Stress Response - ATPase Family AAA Domain-Containing Protein 5
+        "NR-AR-LBD",  # Nuclear Receptor - Androgen Receptor (Ligand Binding Domain)
+    ]
+
+    # Convert assay columns to numeric
+    df[assay_cols] = df[assay_cols].apply(pd.to_numeric, errors="coerce")
+
+    # Create toxic_any column (1 if any assay is positive)
+    df["toxic_any"] = (df[assay_cols] == 1).any(axis=1).astype(int)
+    print(f"Toxicity distribution:\n{df['toxic_any'].value_counts(dropna=False)}")
+
+    # Ensure we have SMILES column (might already be there from LoadSDF)
+    if "smiles" not in df.columns and "molecule" in df.columns:
+        df["smiles"] = df["molecule"].apply(lambda mol: Chem.MolToSmiles(mol) if mol else None)
+
+    # Apply the new tagging system
+    # Use all tag categories by default
+    df = tag_molecules(
+        df,
+        smiles_column="smiles",
+        tag_column="tags",
+        tag_categories=None,  # This will include all: metals, halogens, druglike, structure
+    )
+
+    # Add toxicity-specific tags
+    # Add "tox21" tag for toxic compounds
+    df["tags"] = df.apply(lambda row: row["tags"] + ["tox21_toxic"] if row["toxic_any"] == 1 else row["tags"], axis=1)
+
+    # Add "tox21_clean" tag for non-toxic compounds with data
+    df["tags"] = df.apply(
+        lambda row: (
+            row["tags"] + ["tox21_clean"]
+            if row["toxic_any"] == 0 and not df.loc[row.name, assay_cols].isna().all()
+            else row["tags"]
+        ),
+        axis=1,
+    )
+
+    # Add specific assay tags for positive results
+    for assay in assay_cols:
+        assay_tag = f"positive_{assay.lower().replace('-', '_')}"
+        df["tags"] = df.apply(lambda row: row["tags"] + [assay_tag] if row[assay] == 1 else row["tags"], axis=1)
+
+    # Print tag summary
+    print("\nTag Summary:")
+    tag_summary = get_tag_summary(df)
+    print(tag_summary.head(15))
+
+    # Compute Fingerprints (assuming these functions exist somewhere)
+    df = compute_morgan_fingerprints(df, radius=2)
+    df = project_fingerprints(df, projection="UMAP")
+
+    # Drop the molecule column to save space
+    if "molecule" in df.columns:
+        df.drop(columns=["molecule"], inplace=True)
+
+    # Print final summary
+    print(f"\nFinal DataFrame shape: {df.shape}")
+    print(f"Columns: {df.columns.tolist()}")
+
+    # Show examples of different molecule types
+    print("\n=== Example molecules by category ===")
+
+    # Drug-like molecules
+    druglike = filter_by_tags(df, require=["ro5_pass"])
+    print(f"Drug-like molecules (Ro5 pass): {len(druglike)}")
+
+    # Clean, drug-like, non-toxic molecules
+    ideal = filter_by_tags(df, require=["ro5_pass", "tox21_clean"], exclude=["heavy_metal", "highly_halogenated"])
+    print(f"Ideal molecules (drug-like, non-toxic, no metals/heavy halogenation): {len(ideal)}")
+
+    # Problematic molecules
+    problematic = filter_by_tags(df, require=["tox21_toxic"])
+    print(f"Toxic molecules: {len(problematic)}")
+
+    return df
+
+
+def analyze_toxicity_patterns(df: pd.DataFrame) -> None:
+    """
+    Analyze patterns in toxicity data using the tag system.
+    """
+    print("\n=== Toxicity Pattern Analysis ===")
+
+    # Check correlation between structural features and toxicity
+    toxic_df = filter_by_tags(df, require=["tox21_toxic"])
+    clean_df = filter_by_tags(df, require=["tox21_clean"])
+
+    print(f"\nTotal toxic compounds: {len(toxic_df)}")
+    print(f"Total clean compounds: {len(clean_df)}")
+
+    # Analyze tag distribution in toxic vs clean
+    if len(toxic_df) > 0:
+        print("\nTop tags in toxic compounds:")
+        toxic_tags = get_tag_summary(toxic_df)
+        # Remove the tox21 tags themselves from analysis
+        toxic_tags = toxic_tags[~toxic_tags.index.str.startswith("tox21_")]
+        toxic_tags = toxic_tags[~toxic_tags.index.str.startswith("positive_")]
+        print(toxic_tags.head(10))
+
+    if len(clean_df) > 0:
+        print("\nTop tags in clean compounds:")
+        clean_tags = get_tag_summary(clean_df)
+        clean_tags = clean_tags[~clean_tags.index.str.startswith("tox21_")]
+        clean_tags = clean_tags[~clean_tags.index.str.startswith("positive_")]
+        print(clean_tags.head(10))
+
+    # Check specific patterns
+    print("\n=== Specific Pattern Analysis ===")
+
+    # Heavy metals and toxicity
+    metal_toxic = filter_by_tags(toxic_df, require=["heavy_metal"])
+    print(f"Toxic compounds with heavy metals: {len(metal_toxic)} ({len(metal_toxic) / len(toxic_df) * 100:.1f}%)")
+
+    # Halogenation and toxicity
+    halogen_toxic = filter_by_tags(toxic_df, require=["highly_halogenated"])
+    print(f"Toxic compounds highly halogenated: {len(halogen_toxic)} ({len(halogen_toxic) / len(toxic_df) * 100:.1f}%)")
+
+    # Drug-likeness in toxic compounds
+    druglike_toxic = filter_by_tags(toxic_df, require=["ro5_pass"])
+    print(f"Toxic compounds passing Ro5: {len(druglike_toxic)} ({len(druglike_toxic) / len(toxic_df) * 100:.1f}%)")
+
+
+if __name__ == "__main__":
+    # Load the training and test sets
+    training = "/Users/briford/data/workbench/tox21/training/tox21_10k_data_all.sdf"
+    test = "/Users/briford/data/workbench/tox21/testing/tox21_10k_challenge_test.sdf"
+    final_test = "/Users/briford/data/workbench/tox21/testing/tox21_10k_challenge_score.sdf"
+
+    # Process the training data
+    print("Processing training data...")
+    df = prep_sdf_file(training)
+
+    # Analyze toxicity patterns
+    analyze_toxicity_patterns(df)
+
+    # Store in DFStore
+    df_store = DFStore()
+    df_store.upsert("/datasets/chem_info/tox21_training", df)
+    print(f"\nStored training data to /datasets/chem_info/tox21_training")
+
+    # Optionally process test data
+    # print("\nProcessing test data...")
+    # test_df = prep_sdf_file(test)
+    # df_store.upsert("/datasets/chem_info/tox21_test", test_df)

{workbench-0.8.172 → workbench-0.8.174}/src/workbench/algorithms/graph/light/proximity_graph.py

@@ -135,7 +135,8 @@ if __name__ == "__main__":
     from workbench.algorithms.dataframe.fingerprint_proximity import FingerprintProximity
     from workbench.web_interface.components.plugins.graph_plot import GraphPlot
     from workbench.api import DFStore
-    from workbench.utils.chem_utils import compute_morgan_fingerprints, project_fingerprints
+    from workbench.utils.chem_utils.fingerprints import compute_morgan_fingerprints
+    from workbench.utils.chem_utils.projections import project_fingerprints
     from workbench.utils.graph_utils import connected_sample, graph_layout
 
     def show_graph(graph, id_column):

{workbench-0.8.172 → workbench-0.8.174}/src/workbench/api/compound.py

@@ -3,7 +3,7 @@ import logging
 from typing import List
 
 # Workbench Imports
-from workbench.utils.chem_utils import svg_from_smiles
+from workbench.utils.chem_utils.vis import svg_from_smiles
 
 
 @dataclass

{workbench-0.8.172 → workbench-0.8.174}/src/workbench/api/monitor.py

@@ -15,7 +15,7 @@ class Monitor(MonitorCore):
 
     Common Usage:
        ```
-       mon = Endpoint(name).get_monitor()  # Pull from endpoint OR
+       mon = Endpoint(name).monitor()  # Pull from endpoint OR
        mon = Monitor(name)                 # Create using Endpoint Name
        mon.summary()
        mon.details()
@@ -29,7 +29,6 @@ class Monitor(MonitorCore):
        baseline_df = mon.get_baseline()
        constraints_df = mon.get_constraints()
        stats_df = mon.get_statistics()
-       input_df, output_df = mon.get_captured_data()
        ```
     """
 
@@ -81,15 +80,6 @@ class Monitor(MonitorCore):
         """
         super().create_monitoring_schedule(schedule)
 
-    def get_captured_data(self) -> (pd.DataFrame, pd.DataFrame):
-        """
-        Get the latest data capture input and output from S3.
-
-        Returns:
-            DataFrame (input), DataFrame(output): Flattened and processed DataFrames for input and output data.
-        """
-        return super().get_captured_data()
-
     def get_baseline(self) -> Union[pd.DataFrame, None]:
         """Code to get the baseline CSV from the S3 baseline directory
 
@@ -155,8 +145,3 @@ if __name__ == "__main__":
 
     print("\nStatistics...")
     print(mm.get_statistics())
-
-    # Get the latest data capture
-    input_df, output_df = mm.get_captured_data()
-    print(input_df.head())
-    print(output_df.head())