enzymetk 0.0.2__tar.gz → 0.0.6__tar.gz

{enzymetk-0.0.2 → enzymetk-0.0.6}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: enzymetk
-Version: 0.0.2
+Version: 0.0.6
 Home-page: https://github.com/arianemora/enzyme-tk/
 Author: Ariane Mora
 Author-email: ariane.n.mora@gmail.com
@@ -18,17 +18,12 @@ Classifier: Topic :: Scientific/Engineering :: Bio-Informatics
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: fair-esm
 Requires-Dist: scikit-learn
 Requires-Dist: numpy
 Requires-Dist: seaborn
 Requires-Dist: sciutil
-Requires-Dist: pandas==2.1.4
+Requires-Dist: pandas
 Requires-Dist: biopython
-Requires-Dist: sentence_transformers
-Requires-Dist: pubchempy
-Requires-Dist: pyfaidx
-Requires-Dist: spacy
 Dynamic: author
 Dynamic: author-email
 Dynamic: classifier
@@ -37,6 +32,7 @@ Dynamic: description-content-type
 Dynamic: home-page
 Dynamic: keywords
 Dynamic: license
+Dynamic: license-file
 Dynamic: project-url
 Dynamic: requires-dist
 Dynamic: requires-python
@@ -45,6 +41,9 @@ Dynamic: requires-python
 
 Enzyme-tk is a collection of tools for enzyme engineering, setup as interoperable modules that act on dataframes. These modules are designed to be imported into pipelines for specific function. For this reason, `steps` as each module is called (e.g. finding similar proteins with `BLAST` would be considered a step) are designed to be as light as possible. An example of a pipeline is the [annotate-e](https://github.com/ArianeMora/annotate-e)  ` pipeline, this acts to annotate a fasta with an ensemble of methods (each is designated as an Enzyme-tk step). 
 
+
+**If you have any issues installing, let me know - this has been tested only on Linux/Ubuntu. Please post an issue!**
+
 ## Installation
 
 ## Install base package to import modules
@@ -71,6 +70,7 @@ This is a work-in progress! e.g. some tools (e.g. proteInfer and CLEAN) require
 
 Here are some of the tools that have been implemented to be chained together as a pipeline:
 
+[boltz2](https://github.com/jwohlwend/boltz)
 [mmseqs2](https://github.com/soedinglab/mmseqs2)  
 [foldseek](https://github.com/steineggerlab/foldseek)  
 [diamond](https://github.com/bbuchfink/diamond)  
@@ -89,6 +89,7 @@ Here are some of the tools that have been implemented to be chained together as
 [fasttree](https://morgannprice.github.io/fasttree/)  
 [Porechop](https://github.com/rrwick/Porechop)  
 [prokka](https://github.com/tseemann/prokka)  
+
 ## Things to note
 
 All the tools use the conda env of `enzymetk` by default.
@@ -120,6 +121,8 @@ The steps are the main building blocks of the pipeline. They are responsible for
 
 BLAST is a tool for searching a database of sequences for similar sequences. Here you can either pass a database that you have already created or pass the sequences as part of your dataframe and pass the label column (this needs to have two values: reference and query) reference refers to sequences that you want to search against and query refers to sequences that you want to search for.
 
+Note you need to have installed the BLAST environment.
+
 ```python
 id_col = 'Entry'
 seq_col = 'Sequence'
@@ -148,6 +151,34 @@ df = pd.DataFrame(rows, columns=[id_col, seq_col])
 print(df)
 df << (ActiveSitePred(id_col, seq_col, squidly_dir, num_threads) >> Save('tmp/squidly_as_pred.pkl'))
 
+```
+### Boltz2
+
+Boltz2 is a model for predicting structures. Note you need docko installed as I run via that.
+
+Below is an example using boltz with 4 threads, and uses a cofactor (intermediate in this case). Just set to be None for a single substrate version.
+```
+import sys
+from enzymetk.dock_boltz_step import Boltz
+from enzymetk.save_step import Save
+import pandas as pd
+import os
+os.environ['MKL_THREADING_LAYER'] = 'GNU'
+
+output_dir = 'tmp/'
+num_threads = 4
+id_col = 'Entry'
+seq_col = 'Sequence'
+substrate_col = 'Substrate'
+intermediate_col = 'Intermediate'
+
+rows = [['P0DP23_boltz_8999', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'], 
+        ['P0DP24_boltz_p1', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'],
+        ['P0DP23_boltz_p2', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'], 
+        ['P0DP24_boltz_p3', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'],
+        ['P0DP24_boltz_p4', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]']]
+df = pd.DataFrame(rows, columns=[id_col, seq_col, substrate_col, intermediate_col])
+df << (Boltz(id_col, seq_col, substrate_col, intermediate_col, f'{output_dir}', num_threads) >> Save(f'{output_dir}test.pkl'))
 ```
 
 ### Chai

{enzymetk-0.0.2 → enzymetk-0.0.6}/README.md

@@ -2,6 +2,9 @@
 
 Enzyme-tk is a collection of tools for enzyme engineering, setup as interoperable modules that act on dataframes. These modules are designed to be imported into pipelines for specific function. For this reason, `steps` as each module is called (e.g. finding similar proteins with `BLAST` would be considered a step) are designed to be as light as possible. An example of a pipeline is the [annotate-e](https://github.com/ArianeMora/annotate-e)  ` pipeline, this acts to annotate a fasta with an ensemble of methods (each is designated as an Enzyme-tk step). 
 
+
+**If you have any issues installing, let me know - this has been tested only on Linux/Ubuntu. Please post an issue!**
+
 ## Installation
 
 ## Install base package to import modules
@@ -28,6 +31,7 @@ This is a work-in progress! e.g. some tools (e.g. proteInfer and CLEAN) require
 
 Here are some of the tools that have been implemented to be chained together as a pipeline:
 
+[boltz2](https://github.com/jwohlwend/boltz)
 [mmseqs2](https://github.com/soedinglab/mmseqs2)  
 [foldseek](https://github.com/steineggerlab/foldseek)  
 [diamond](https://github.com/bbuchfink/diamond)  
@@ -46,6 +50,7 @@ Here are some of the tools that have been implemented to be chained together as
 [fasttree](https://morgannprice.github.io/fasttree/)  
 [Porechop](https://github.com/rrwick/Porechop)  
 [prokka](https://github.com/tseemann/prokka)  
+
 ## Things to note
 
 All the tools use the conda env of `enzymetk` by default.
@@ -77,6 +82,8 @@ The steps are the main building blocks of the pipeline. They are responsible for
 
 BLAST is a tool for searching a database of sequences for similar sequences. Here you can either pass a database that you have already created or pass the sequences as part of your dataframe and pass the label column (this needs to have two values: reference and query) reference refers to sequences that you want to search against and query refers to sequences that you want to search for.
 
+Note you need to have installed the BLAST environment.
+
 ```python
 id_col = 'Entry'
 seq_col = 'Sequence'
@@ -105,6 +112,34 @@ df = pd.DataFrame(rows, columns=[id_col, seq_col])
 print(df)
 df << (ActiveSitePred(id_col, seq_col, squidly_dir, num_threads) >> Save('tmp/squidly_as_pred.pkl'))
 
+```
+### Boltz2
+
+Boltz2 is a model for predicting structures. Note you need docko installed as I run via that.
+
+Below is an example using boltz with 4 threads, and uses a cofactor (intermediate in this case). Just set to be None for a single substrate version.
+```
+import sys
+from enzymetk.dock_boltz_step import Boltz
+from enzymetk.save_step import Save
+import pandas as pd
+import os
+os.environ['MKL_THREADING_LAYER'] = 'GNU'
+
+output_dir = 'tmp/'
+num_threads = 4
+id_col = 'Entry'
+seq_col = 'Sequence'
+substrate_col = 'Substrate'
+intermediate_col = 'Intermediate'
+
+rows = [['P0DP23_boltz_8999', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'], 
+        ['P0DP24_boltz_p1', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'],
+        ['P0DP23_boltz_p2', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'], 
+        ['P0DP24_boltz_p3', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'],
+        ['P0DP24_boltz_p4', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]']]
+df = pd.DataFrame(rows, columns=[id_col, seq_col, substrate_col, intermediate_col])
+df << (Boltz(id_col, seq_col, substrate_col, intermediate_col, f'{output_dir}', num_threads) >> Save(f'{output_dir}test.pkl'))
 ```
 
 ### Chai

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/__init__.py

@@ -22,34 +22,11 @@ Date: March 2025
 __title__ = 'enzymetk'
 __description__ = 'Toolkit for enzymes and what not'
 __url__ = 'https://github.com/arianemora/enzyme-tk/'
-__version__ = '0.0.2'
+__version__ = '0.0.6'
 __author__ = 'Ariane Mora'
 __author_email__ = 'ariane.n.mora@gmail.com'
 __license__ = 'GPL3'
 
-# from enzymetk.step import *
-# from enzymetk.generate_msa_step import ClustalOmega
-# from enzymetk.annotateEC_CLEAN_step import CLEAN
-# from enzymetk.annotateEC_proteinfer_step import ProteInfer
-# from enzymetk.dock_chai_step import Chai
-# from enzymetk.dock_vina_step import Vina
-# from enzymetk.embedchem_chemberta_step import ChemBERT
-# from enzymetk.embedchem_rxnfp_step import RxnFP
-# from enzymetk.embedchem_selformer_step import SelFormer
-# from enzymetk.embedchem_unimol_step import UniMol
-# from enzymetk.embedprotein_esm_step import EmbedESM
-# from enzymetk.generate_tree_step import FastTree
-# from enzymetk.inpaint_ligandMPNN_step import LigandMPNN
-# from enzymetk.metagenomics_porechop_trim_reads_step import PoreChop
-# from enzymetk.metagenomics_prokka_annotate_genes import Prokka
-# #from enzymetk.predict_activity_step import 
-# from enzymetk.predict_catalyticsite_step import ActiveSitePred
-# from enzymetk.sequence_search_blast import BLAST
-# from enzymetk.similarity_foldseek_step import FoldSeek
-# from enzymetk.similarity_mmseqs_step import MMseqs
-# from enzymetk.similarity_reaction_step import ReactionDist
-# from enzymetk.similarity_substrate_step import SubstrateDist
-
 
 
 

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/annotateEC_CLEAN_step.py

@@ -116,7 +116,7 @@ class CLEAN(Step):
                 print(output_filenames)
                 for sub_df in output_filenames:
                     df = pd.concat([df, sub_df])
-                return df
+                return self.__filter_df(df)
             else:
-                return self.__execute([df, tmp_dir])
+                return self.__filter_df(self.__execute([df, tmp_dir]))
                 return df

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/annotateEC_CREEP_step.py

@@ -38,7 +38,7 @@ class CREEP(Step):
         self.args_extract = args_extract
         self.args_retrieval = args_retrieval
             
-    def __execute(self, df: pd.DataFrame, tmp_dir: str) -> pd.DataFrame:
+    def __execute(self, df: pd.DataFrame, tmp_dir: str):
         tmp_dir = '/disk1/ariane/vscode/degradeo/pipeline/tmp/'
         input_filename = f'{tmp_dir}/creepasjkdkajshdkja.csv'
         df.to_csv(input_filename, index=False)

enzymetk-0.0.6/enzymetk/dock_boltz_step.py

@@ -0,0 +1,46 @@
+from enzymetk.step import Step
+import pandas as pd
+from docko.boltz import run_boltz_affinity
+import logging
+import numpy as np
+from multiprocessing.dummy import Pool as ThreadPool
+
+
+logger = logging.getLogger(__name__)
+logger.setLevel(logging.INFO)
+
+    
+class Boltz(Step):
+    
+    def __init__(self, id_col: str, seq_col: str, substrate_col: str, intermediate_col: str, output_dir: str, num_threads: int):
+        self.id_col = id_col
+        self.seq_col = seq_col
+        self.substrate_col = substrate_col
+        self.intermediate_col = intermediate_col
+        self.output_dir = output_dir or None
+        self.num_threads = num_threads or 1
+
+    def __execute(self, df: pd.DataFrame) -> pd.DataFrame:
+        output_filenames = []
+        
+        for run_id, seq, substrate, intermediate in df[[self.id_col, self.seq_col, self.substrate_col, self.intermediate_col]].values:
+            # Might have an issue if the things are not correctly installed in the same dicrectory 
+            if not isinstance(substrate, str):
+                substrate = ''
+            print(run_id, seq, substrate)
+            run_boltz_affinity(run_id, seq, substrate, self.output_dir, intermediate)
+            output_filenames.append(f'{self.output_dir}/{run_id}/')
+        return output_filenames
+    
+    def execute(self, df: pd.DataFrame) -> pd.DataFrame:
+        if self.output_dir:
+            if self.num_threads > 1:
+                pool = ThreadPool(self.num_threads)
+                df_list = np.array_split(df, self.num_threads)
+                results = pool.map(self.__execute, df_list)
+            else:
+                results = self.__execute(df)
+            df['output_dir'] = results
+            return df
+        else:
+            print('No output directory provided')

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/dock_chai_step.py

@@ -11,16 +11,17 @@ logger.setLevel(logging.INFO)
     
 class Chai(Step):
     
-    def __init__(self, id_col: str, seq_col: str, substrate_col: str, output_dir: str, num_threads: int):
+    def __init__(self, id_col: str, seq_col: str, substrate_col: str, cofactor_col: str, output_dir: str, num_threads: int):
         self.id_col = id_col
         self.seq_col = seq_col
         self.substrate_col = substrate_col
+        self.cofactor_col = cofactor_col
         self.output_dir = output_dir or None
         self.num_threads = num_threads or 1
 
     def __execute(self, df: pd.DataFrame, tmp_dir: str) -> pd.DataFrame:
         output_filenames = []
-        for run_id, seq, substrate in df[[self.id_col, self.seq_col, self.substrate_col]].values:
+        for run_id, seq, substrate, cofactor in df[[self.id_col, self.seq_col, self.substrate_col, self.cofactor_col]].values:
             # Might have an issue if the things are not correctly installed in the same dicrectory 
             if not isinstance(substrate, str):
                 substrate = ''
@@ -28,7 +29,8 @@ class Chai(Step):
             run_chai(run_id, # name
                     seq, # sequence
                     substrate, # ligand as smiles
-                    tmp_dir)
+                    tmp_dir,
+                    cofactor) # cofactor as smiles
             output_filenames.append(f'{tmp_dir}/{run_id}/')
         return output_filenames
     

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/dock_vina_step.py

@@ -4,6 +4,7 @@ from docko.docko import *
 import logging
 import numpy as np
 import os
+from pathlib import Path
 from multiprocessing.dummy import Pool as ThreadPool
 
 logger = logging.getLogger(__name__)
@@ -21,34 +22,66 @@ class Vina(Step):
         self.substrate_col = substrate_col
         self.substrate_name_col = substrate_name_col
         self.active_site_col = active_site_col  # Expects active site residues as a string separated by |
-        self.output_dir = output_dir or None
+        self.output_dir = Path( output_dir) or None
         self.num_threads = num_threads or 1
 
     def __execute(self, df: pd.DataFrame) -> pd.DataFrame:
         output_filenames = []
         # ToDo: update to create from sequence if the path doesn't exist.
         for label, structure_path, seq, substrate_smiles, substrate_name, residues in df[[self.id_col, self.structure_col, self.sequence_col, self.substrate_col, self.substrate_name_col, self.active_site_col]].values:
-            os.system(f'mkdir {self.output_dir}{label}')
+
             try:
+                structure_path = str(structure_path)
                 residues = str(residues)
                 residues = [int(r) + 1 for r in residues.split('|')]
-                if not os.path.exists(f'{structure_path}'):
-                    # Try get the AF2 structure we expect the label to be the uniprot id
-                    get_alphafold_structure(label, f'{self.output_dir}{label}/{label}_AF2.pdb')
-                    structure_path = f'{self.output_dir}{label}/{label}_AF2.pdb'
-                clean_one_pdb(f'{structure_path}', f'{self.output_dir}{label}/{label}.pdb')
-                pdb_to_pdbqt_protein(f'{self.output_dir}{label}/{label}.pdb', f'{self.output_dir}{label}/{label}.pdbqt')
-                score = dock(sequence='', protein_name=label, smiles=substrate_smiles, ligand_name=substrate_name, residues=residues, 
-                            protein_dir=f'{self.output_dir}', ligand_dir=f'{self.output_dir}', output_dir=f'{self.output_dir}{label}/', pH=7.4,
-                            method='vina', size_x=10.0, size_y=10.0, size_z=10.0)
-                output_filename = f'{self.output_dir}{label}/{label}.pdb'
-                output_filenames.append(output_filename)
+
+                label_dir = self.output_dir / label
+                label_dir.mkdir(parents=True, exist_ok=True)
+                structure_path = Path(structure_path)
+
+                if not structure_path.exists():
+                    # Try to download AF2 structure
+                    get_alphafold_structure(label, label_dir / f"{label}_AF2.pdb")
+                    structure_path = label_dir / f"{label}_AF2.pdb"
+
+                # Skip if still not found
+                if not structure_path.exists():
+                    print(f"Skipping {label}: AF2 structure not found.")
+                    output_filenames.append(None)
+                    continue
+            
+                # Proceed with docking
+                pdb_path = label_dir / f"{label}.pdb"
+                pdbqt_path = label_dir / f"{label}.pdbqt"
+
+                clean_one_pdb(str(structure_path), str(pdb_path))
+                pdb_to_pdbqt_protein(str(pdb_path), str(pdbqt_path))
+
+                score = dock(
+                    sequence='',
+                    protein_name=label,
+                    smiles=substrate_smiles,
+                    ligand_name=substrate_name,
+                    residues=residues,
+                    protein_dir=str(self.output_dir),
+                    ligand_dir=str(self.output_dir),
+                    output_dir=str(label_dir),
+                    pH=7.4,
+                    method='vina',
+                    size_x=10.0,
+                    size_y=10.0,
+                    size_z=10.0
+                )
+
+                output_filenames.append(str(pdb_path))
+                
             except Exception as e:
                 print(f'Error docking {label}: {e}')
                 output_filenames.append(None)
+            
         return output_filenames
-        
-    
+
+ 
     def execute(self, df: pd.DataFrame) -> pd.DataFrame:
         if self.output_dir:
             if self.num_threads > 1:
@@ -60,4 +93,4 @@ class Vina(Step):
             df['output_dir'] = results
             return df
         else:
-            print('No output directory provided')
+            print('No output directory provided')

enzymetk-0.0.6/enzymetk/embedprotein_esm3_step.py

@@ -0,0 +1,71 @@
+# ESM 3 script
+from esm.sdk.api import ESMProtein
+from tempfile import TemporaryDirectory
+import torch
+import os
+import pandas as pd
+from esm.models.esm3 import ESM3
+from esm.sdk.api import ESMProtein, SamplingConfig
+from huggingface_hub import login
+from enzymetk.step import Step
+import numpy as np
+from tqdm import tqdm 
+
+# CUDA setup
+os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"   # see issue #152
+cuda = True
+DEVICE = torch.device("cuda" if cuda else "cpu")
+device = DEVICE 
+
+
+class EmbedESM3(Step):
+    
+    def __init__(self, id_col: str, seq_col: str, extraction_method='mean', num_threads=1, 
+                 tmp_dir: str = None, env_name: str = 'enzymetk', save_tensors=False): # type: ignore
+        login()
+        self.client = ESM3.from_pretrained("esm3-open").to("cuda")
+        self.seq_col = seq_col
+        self.id_col = id_col
+        self.num_threads = num_threads or 1
+        self.extraction_method = extraction_method
+        self.tmp_dir = tmp_dir
+        self.env_name = env_name
+        self.save_tensors = save_tensors
+
+    def __execute(self, df: pd.DataFrame, tmp_dir: str) -> pd.DataFrame: 
+        client = self.client
+        means = []
+        for id, seq in tqdm(df[[self.id_col, self.seq_col]].values):
+            protein = ESMProtein(
+                sequence=(
+                    seq
+                )
+            )
+            protein_tensor = client.encode(protein)
+            output = client.forward_and_sample(
+                protein_tensor, SamplingConfig(return_per_residue_embeddings=True)
+            )
+            if self.save_tensors:
+                torch.save(output.per_residue_embedding, os.path.join(tmp_dir, f'{id}.pt'))
+            means.append(np.array(output.per_residue_embedding.mean(dim=0).cpu()))
+        df['esm3_mean']  = means
+        return df
+    
+    def execute(self, df: pd.DataFrame) -> pd.DataFrame:
+        if self.tmp_dir is None:
+            with TemporaryDirectory() as tmp_dir:
+                if self.num_threads > 1:
+                    dfs = []
+                    df_list = np.array_split(df, self.num_threads)
+                    for df_chunk in tqdm(df_list):
+                        dfs.append(self.__execute(df_chunk, tmp_dir))
+                    df = pd.DataFrame()
+                    for tmp_df in tqdm(dfs):
+                        df = pd.concat([df, tmp_df])
+                    return df
+                else:
+                    df = self.__execute(df, tmp_dir)
+                    return df
+        else:
+            df = self.__execute(df, self.tmp_dir)
+            return df

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/embedprotein_esm_step.py

@@ -70,8 +70,8 @@ def extract_mean_embedding(df, id_column, encoding_dir, rep_num=33):
 
 class EmbedESM(Step):
     
-    def __init__(self, id_col: str, seq_col: str, model='esm2_t33_650M_UR50D', extraction_method='mean', 
-                 active_site_col: str = None, num_threads=1, tmp_dir: str = None, env_name: str = 'enzymetk'):
+    def __init__(self, id_col: str, seq_col: str, model='esm2_t36_3B_UR50D', extraction_method='mean', 
+                 active_site_col: str = None, num_threads=1, tmp_dir: str = None, env_name: str = 'enzymetk', rep_num=36):
         self.seq_col = seq_col
         self.id_col = id_col
         self.active_site_col = active_site_col
@@ -80,6 +80,7 @@ class EmbedESM(Step):
         self.extraction_method = extraction_method
         self.tmp_dir = tmp_dir
         self.env_name = env_name
+        self.rep_num = rep_num
 
     def __execute(self, df: pd.DataFrame, tmp_dir: str) -> pd.DataFrame:
         input_filename = f'{tmp_dir}/input.fasta'
@@ -95,11 +96,11 @@ class EmbedESM(Step):
         cmd = ['conda', 'run', '-n', self.env_name, 'python', Path(__file__).parent/'esm-extract.py', self.model, input_filename, tmp_dir, '--include', 'per_tok']
         self.run(cmd)
         if self.extraction_method == 'mean':
-            df = extract_mean_embedding(df, self.id_col, tmp_dir)
+            df = extract_mean_embedding(df, self.id_col, tmp_dir, rep_num=self.rep_num)
         elif self.extraction_method == 'active_site':
             if self.active_site_col is None:
                 raise ValueError('active_site_col must be provided if extraction_method is active_site')
-            df = extract_active_site_embedding(df, self.id_col, self.active_site_col, tmp_dir)
+            df = extract_active_site_embedding(df, self.id_col, self.active_site_col, tmp_dir, rep_num=self.rep_num)
         
         return df
     

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/inpaint_ligandMPNN_step.py

@@ -16,7 +16,8 @@ import os
 
 class LigandMPNN(Step):
     
-    def __init__(self, pdb_column_name: str, ligand_mpnn_dir: str, output_dir: str, tmp_dir: str = None, args=None, num_threads: int = 1, env_name: str = 'ligandmpnn_env'):
+    def __init__(self, pdb_column_name: str, ligand_mpnn_dir: str, output_dir: str, 
+                 tmp_dir: str = None, args=None, num_threads: int = 1, env_name: str = 'ligandmpnn_env'):
         self.pdb_column_name = pdb_column_name
         self.ligand_mpnn_dir = ligand_mpnn_dir
         self.output_dir = output_dir
@@ -34,7 +35,7 @@ class LigandMPNN(Step):
         os.chdir(self.ligand_mpnn_dir)
         
         for pdb_file in df[ self.pdb_column_name].values:
-            cmd = ['conda', 'run', '-n', self.env_name, 'python3', f'{self.ligand_mpnn_dir}run.py', '--pdb_path', pdb_file,  '--out_folder', f'{self.output_dir}']
+            cmd = ['conda', 'run', '-n', self.env_name, 'python3', f'{self.ligand_mpnn_dir}run.py', '--pdb_path', pdb_file,  '--out_folder', f'{self.output_dir}'] 
             if self.args is not None:
                 cmd.extend(self.args)
             result = subprocess.run(cmd, check=True)

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/predict_catalyticsite_step.py

@@ -8,6 +8,8 @@ import numpy as np
 from tqdm import tqdm
 import random
 import string
+import logging
+import os
 
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.INFO)
@@ -15,15 +17,17 @@ logger.setLevel(logging.INFO)
     
 class ActiveSitePred(Step):
     
-    def __init__(self, id_col: str, seq_col: str, squidly_dir: str, num_threads: int = 1, 
-                 esm2_model = 'esm2_t36_3B_UR50D', tmp_dir: str = None):
+    def __init__(self, id_col: str, seq_col: str, num_threads: int = 1, 
+                 esm2_model = 'esm2_t36_3B_UR50D', tmp_dir: str = None, args=None):
         self.id_col = id_col
         self.seq_col = seq_col  
         self.num_threads = num_threads or 1
-        self.squidly_dir = squidly_dir
         self.esm2_model = esm2_model
         self.tmp_dir = tmp_dir
-
+        self.args = None
+        self.logger = logging.getLogger(__name__)
+        print('Predicting Active Sites using Squidly')
+        
     def __to_fasta(self, df: pd.DataFrame, tmp_dir: str):
         tmp_label = ''.join(random.choices(string.ascii_letters + string.digits, k=10))
 
@@ -37,13 +41,17 @@ class ActiveSitePred(Step):
     def __execute(self, df: pd.DataFrame, tmp_dir: str):
         input_filename = self.__to_fasta(df, tmp_dir)
         # Might have an issue if the things are not correctly installed in the same dicrectory 
-        result = subprocess.run(['python', Path(__file__).parent/'predict_catalyticsite_run.py', '--out', str(tmp_dir), 
-                                '--input', input_filename, '--squidly_dir', self.squidly_dir, '--esm2_model', self.esm2_model], capture_output=True, text=True)
-        output_filename = f'{input_filename.replace(".fasta", "_results.pkl")}'
+        cmd = []
+        cmd = ['squidly', 'run', input_filename, self.esm2_model, tmp_dir]
+        if self.args is not None:
+            cmd.extend(self.args)
+        result = self.run(cmd)
         if result.stderr:
-            logger.error(result.stderr)
-        logger.info(result.stdout)   
-        
+            self.logger.error(result.stderr)
+            print(result.stderr)
+        else:
+            self.logger.info(result.stdout)
+        output_filename = os.path.join(tmp_dir, 'squidly_ensemble.csv')
         return output_filename
     
     def execute(self, df: pd.DataFrame) -> pd.DataFrame:
@@ -61,10 +69,10 @@ class ActiveSitePred(Step):
                 df = pd.DataFrame()
                 print(output_filenames)
                 for p in output_filenames:
-                    sub_df = pd.read_pickle(p)
+                    sub_df = pd.read_csv(p)
                     df = pd.concat([df, sub_df])
                 return df
             
             else:
                 output_filename = self.__execute(df, tmp_dir)
-                return pd.read_pickle(output_filename)
+                return pd.read_csv(output_filename)

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/similarity_foldseek_step.py

@@ -125,11 +125,10 @@ class FoldSeek(Step):
                          continue
                 df = pd.DataFrame()
                 print(output_filenames)
-                for p in output_filenames:
-                    sub_df = pd.read_pickle(p)
+                for sub_df in output_filenames:
                     df = pd.concat([df, sub_df])
                 return df
             
             else:
-                output_filename = self.__execute([df, tmp_dir])
-                return pd.read_pickle(output_filename)
+                df = self.__execute([df, tmp_dir])
+                return df

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/similarity_reaction_step.py

@@ -24,22 +24,26 @@ class ReactionDist(Step):
         self.num_threads = num_threads
         
     def __execute(self, data: list) -> np.array:
-        reaction_df = data
-        tmp_label = ''.join(random.choices(string.ascii_letters + string.digits, k=10))
-        
-        rxn = rdChemReactions.ReactionFromSmarts(self.smiles_string)
-        rxn_fp = rdChemReactions.CreateStructuralFingerprintForReaction(rxn)
+        reaction_df = data        
         rows = []
+        fp_params = rdChemReactions.ReactionFingerprintParams()
+        rxn = rdChemReactions.ReactionFromSmarts(self.smiles_string)
+        rxn_fp = rdChemReactions.CreateStructuralFingerprintForReaction(rxn, ReactionFingerPrintParams=fp_params) #rdChemReactions.CreateStructuralFingerprintForReaction(rxn, ReactionFingerPrintParams=fp_params)
+
         # compare all fp pairwise without duplicates
         for smile_id, smiles in tqdm(reaction_df[[self.id_column_name, self.smiles_column_name]].values): # -1 so the last fp will not be used
             mol_ = rdChemReactions.ReactionFromSmarts(smiles)
-            fps = rdChemReactions.CreateStructuralFingerprintForReaction(mol_)
-            rows.append([smile_id, 
+            # Note: if you don't pass , ReactionFingerPrintParams=fp_params you get different results
+            # i.e. reactions that don't appear to be the same are reported as similar of 1.0
+            # https://github.com/rdkit/rdkit/discussions/5263
+            fps = rdChemReactions.CreateStructuralFingerprintForReaction(mol_, ReactionFingerPrintParams=fp_params)
+            rows.append([smile_id,
+                         self.smiles_string, 
                          smiles, 
                          DataStructs.TanimotoSimilarity(fps, rxn_fp), 
                          DataStructs.RusselSimilarity(fps, rxn_fp), 
                          DataStructs.CosineSimilarity(fps, rxn_fp)])
-        distance_df = pd.DataFrame(rows, columns=[self.id_column_name, 'TargetSmiles', 'TanimotoSimilarity', 'RusselSimilarity', 'CosineSimilarity'])
+        distance_df = pd.DataFrame(rows, columns=[self.id_column_name, 'QuerySmiles', 'TargetSmiles', 'TanimotoSimilarity', 'RusselSimilarity', 'CosineSimilarity'])
         return distance_df
         
     def execute(self, df: pd.DataFrame) -> pd.DataFrame:

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/similarity_substrate_step.py

@@ -1,3 +1,5 @@
+import sys
+sys.path.append('/disk1/ariane/vscode/enzyme-tk/')
 from enzymetk.step import Step
 import pandas as pd
 import numpy as np
@@ -8,6 +10,7 @@ from rdkit.Chem import rdChemReactions
 import pandas as pd
 import os
 from rdkit.DataStructs import FingerprintSimilarity
+from rdkit.Chem import rdFingerprintGenerator
 from rdkit.Chem.Fingerprints import FingerprintMols
 import random
 import string
@@ -28,12 +31,15 @@ class SubstrateDist(Step):
         tmp_label = ''.join(random.choices(string.ascii_letters + string.digits, k=10))
         
         rxn = Chem.MolFromSmiles(self.smiles_string)
-        rxn_fp = FingerprintMols.FingerprintMol(rxn)
+        # Switched to using morgan fingerprints https://greglandrum.github.io/rdkit-blog/posts/2023-01-18-fingerprint-generator-tutorial.html
+        # followed this tutorial
+        mfpgen = rdFingerprintGenerator.GetMorganGenerator(radius=2,fpSize=2048)
+        rxn_fp = mfpgen.GetFingerprint(rxn)
         rows = []
         # compare all fp pairwise without duplicates
         for smile_id, smiles in tqdm(reaction_df[[self.id_column_name, self.smiles_column_name]].values): # -1 so the last fp will not be used
             mol_ = Chem.MolFromSmiles(smiles)
-            fps = FingerprintMols.FingerprintMol(mol_)
+            fps = mfpgen.GetFingerprint(mol_)
             rows.append([smile_id, 
                          smiles, 
                          DataStructs.TanimotoSimilarity(fps, rxn_fp), 

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk/step.py

@@ -36,8 +36,9 @@ class Step():
         """ Execute some shit """ 
         return df
     
-    def run(self, cmd: list) -> None:
-        """ Run a command """
+    def run(self, cmd: list):
+        """ Run a command """   
+        result = None
         start = timeit.default_timer()
         u.dp(['Running command', ' '.join([str(c) for c in cmd])])
         result = subprocess.run(cmd, capture_output=True, text=True)       
@@ -48,8 +49,9 @@ class Step():
             logger.error(result.stderr)
         logger.info(result.stdout)
         u.dp(['Time for command to run (min): ', (timeit.default_timer() - start)/60])
+        return result
 
-    def __rshift__(self, other: Step) -> Step:
+    def __rshift__(self, other: Step)   :
         return Pipeline(self, other)
         
     def __rlshift__(self, other: pd.DataFrame) -> pd.DataFrame:

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: enzymetk
-Version: 0.0.2
+Version: 0.0.6
 Home-page: https://github.com/arianemora/enzyme-tk/
 Author: Ariane Mora
 Author-email: ariane.n.mora@gmail.com
@@ -18,17 +18,12 @@ Classifier: Topic :: Scientific/Engineering :: Bio-Informatics
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: fair-esm
 Requires-Dist: scikit-learn
 Requires-Dist: numpy
 Requires-Dist: seaborn
 Requires-Dist: sciutil
-Requires-Dist: pandas==2.1.4
+Requires-Dist: pandas
 Requires-Dist: biopython
-Requires-Dist: sentence_transformers
-Requires-Dist: pubchempy
-Requires-Dist: pyfaidx
-Requires-Dist: spacy
 Dynamic: author
 Dynamic: author-email
 Dynamic: classifier
@@ -37,6 +32,7 @@ Dynamic: description-content-type
 Dynamic: home-page
 Dynamic: keywords
 Dynamic: license
+Dynamic: license-file
 Dynamic: project-url
 Dynamic: requires-dist
 Dynamic: requires-python
@@ -45,6 +41,9 @@ Dynamic: requires-python
 
 Enzyme-tk is a collection of tools for enzyme engineering, setup as interoperable modules that act on dataframes. These modules are designed to be imported into pipelines for specific function. For this reason, `steps` as each module is called (e.g. finding similar proteins with `BLAST` would be considered a step) are designed to be as light as possible. An example of a pipeline is the [annotate-e](https://github.com/ArianeMora/annotate-e)  ` pipeline, this acts to annotate a fasta with an ensemble of methods (each is designated as an Enzyme-tk step). 
 
+
+**If you have any issues installing, let me know - this has been tested only on Linux/Ubuntu. Please post an issue!**
+
 ## Installation
 
 ## Install base package to import modules
@@ -71,6 +70,7 @@ This is a work-in progress! e.g. some tools (e.g. proteInfer and CLEAN) require
 
 Here are some of the tools that have been implemented to be chained together as a pipeline:
 
+[boltz2](https://github.com/jwohlwend/boltz)
 [mmseqs2](https://github.com/soedinglab/mmseqs2)  
 [foldseek](https://github.com/steineggerlab/foldseek)  
 [diamond](https://github.com/bbuchfink/diamond)  
@@ -89,6 +89,7 @@ Here are some of the tools that have been implemented to be chained together as
 [fasttree](https://morgannprice.github.io/fasttree/)  
 [Porechop](https://github.com/rrwick/Porechop)  
 [prokka](https://github.com/tseemann/prokka)  
+
 ## Things to note
 
 All the tools use the conda env of `enzymetk` by default.
@@ -120,6 +121,8 @@ The steps are the main building blocks of the pipeline. They are responsible for
 
 BLAST is a tool for searching a database of sequences for similar sequences. Here you can either pass a database that you have already created or pass the sequences as part of your dataframe and pass the label column (this needs to have two values: reference and query) reference refers to sequences that you want to search against and query refers to sequences that you want to search for.
 
+Note you need to have installed the BLAST environment.
+
 ```python
 id_col = 'Entry'
 seq_col = 'Sequence'
@@ -148,6 +151,34 @@ df = pd.DataFrame(rows, columns=[id_col, seq_col])
 print(df)
 df << (ActiveSitePred(id_col, seq_col, squidly_dir, num_threads) >> Save('tmp/squidly_as_pred.pkl'))
 
+```
+### Boltz2
+
+Boltz2 is a model for predicting structures. Note you need docko installed as I run via that.
+
+Below is an example using boltz with 4 threads, and uses a cofactor (intermediate in this case). Just set to be None for a single substrate version.
+```
+import sys
+from enzymetk.dock_boltz_step import Boltz
+from enzymetk.save_step import Save
+import pandas as pd
+import os
+os.environ['MKL_THREADING_LAYER'] = 'GNU'
+
+output_dir = 'tmp/'
+num_threads = 4
+id_col = 'Entry'
+seq_col = 'Sequence'
+substrate_col = 'Substrate'
+intermediate_col = 'Intermediate'
+
+rows = [['P0DP23_boltz_8999', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'], 
+        ['P0DP24_boltz_p1', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'],
+        ['P0DP23_boltz_p2', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'], 
+        ['P0DP24_boltz_p3', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]'],
+        ['P0DP24_boltz_p4', 'MALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAAMALWMRLLPLLALLALWGPDPAAA', 'CCCCC(CC)COC(=O)C1=CC=CC=C1C(=O)OCC(CC)CCCC', 'CC1=C(C2=CC3=C(C(=C([N-]3)C=C4C(=C(C(=N4)C=C5C(=C(C(=N5)C=C1[N-]2)C)C=C)C)C=C)C)CCC(=O)[O-])CCC(=O)[O-].[Fe]']]
+df = pd.DataFrame(rows, columns=[id_col, seq_col, substrate_col, intermediate_col])
+df << (Boltz(id_col, seq_col, substrate_col, intermediate_col, f'{output_dir}', num_threads) >> Save(f'{output_dir}test.pkl'))
 ```
 
 ### Chai

{enzymetk-0.0.2 → enzymetk-0.0.6}/enzymetk.egg-info/SOURCES.txt

@@ -5,6 +5,7 @@ enzymetk/__init__.py
 enzymetk/annotateEC_CLEAN_step.py
 enzymetk/annotateEC_CREEP_step.py
 enzymetk/annotateEC_proteinfer_step.py
+enzymetk/dock_boltz_step.py
 enzymetk/dock_chai_step.py
 enzymetk/dock_vina_step.py
 enzymetk/embedchem_chemberta_step.py
@@ -13,6 +14,7 @@ enzymetk/embedchem_rxnfp_step.py
 enzymetk/embedchem_selformer_run.py
 enzymetk/embedchem_selformer_step.py
 enzymetk/embedchem_unimol_step.py
+enzymetk/embedprotein_esm3_step.py
 enzymetk/embedprotein_esm_step.py
 enzymetk/esm-extract.py
 enzymetk/filter_sequence_step.py

enzymetk-0.0.6/enzymetk.egg-info/requires.txt

@@ -0,0 +1,6 @@
+scikit-learn
+numpy
+seaborn
+sciutil
+pandas
+biopython

{enzymetk-0.0.2 → enzymetk-0.0.6}/setup.py

@@ -61,17 +61,13 @@ setup(name='enzymetk',
               'enzymetk = enzymetk.__main__:main'
           ]
       },
-      install_requires=['fair-esm',
+      install_requires=[
                         'scikit-learn',
                         'numpy',
                         'seaborn',
                         'sciutil',
-                        'pandas==2.1.4',
-                        'biopython',
-                        'sentence_transformers',
-                        'pubchempy',
-                        'pyfaidx',
-                        'spacy'],
+                        'pandas',
+                        'biopython'],
       python_requires='>=3.8',
       data_files=[("", ["LICENSE"])]
       )

enzymetk-0.0.2/enzymetk.egg-info/requires.txt

@@ -1,11 +0,0 @@
-fair-esm
-scikit-learn
-numpy
-seaborn
-sciutil
-pandas==2.1.4
-biopython
-sentence_transformers
-pubchempy
-pyfaidx
-spacy