# Copyright (c) 2025 Isom Lab # Released under the MIT License; see LICENSE at the repository root. """ ========================================================================================================= # Script: **Pairwise Structural Alignment with TM-align** # Description: # This script **performs large-scale pairwise structure alignment** of a **query structure** # against **214,528,851 AlphaFold Database predictions**. It distributes computations # across **1000 compute nodes** on the **Triton HPC cluster**. ========================================================================================================= 📌 **CONFIGURATION PARAMETERS** --------------------------------------------------------------------------------------------------------- 🔹 **Cluster Information** - `cluster` → Specifies the cluster (**Triton** or **Pegasus**). - `queue` → Job submission queue (**short** or **normal**). 🔹 **Save Directories** - `save_directory` → Root directory for storing results. - `calculation_name` → Subdirectory for this specific calculation. - `calculation_save_path` → Full path where results are stored. 🔹 **File Locations** - `where_query_pdb_files_are_located` → Directory containing **query PDB structures**. - `where_target_cif_set_files_are_located` → Location of **AlphaFold structure submission sets**. - `superdark_script` → Path to the **alignment script** (`1-superDark.py`). 🔹 **Job Resubmission** - `job_resubmission_file` → If resubmitting failed jobs, this file contains **job numbers**. - `job_resubmission_numbers` → List of **job IDs** to be resubmitted. ========================================================================================================= 📌 **DEPENDENCIES** --------------------------------------------------------------------------------------------------------- - `os`, `shutil`, `subprocess` → Used for **file handling**, **job submission**, **directory creation**, and **monitoring job progress**. ========================================================================================================= 📌 **EXECUTION FLOW** --------------------------------------------------------------------------------------------------------- 🔹 **Step 1: Create Necessary Directories** - **Creates or refreshes** directories for: 1️⃣ **Calculation results** 2️⃣ **Log files** 3️⃣ **Job submission inputs** 🔹 **Step 2: Collect Job Submission Sets** - If **not resubmitting**, retrieves **all** 1000 job set files. - If **resubmitting**, collects **only the failed job set files**. 🔹 **Step 3: Prepare Job Submission Files** - Generates **query structure paths** from the **query directory**. - Reads **target structure paths** from the **submission sets**. - Creates **combined query-target file lists** for TM-align processing. 🔹 **Step 4: Submit Jobs to the Cluster** - **Creates job submission files** (each containing query-target pairs). - Uses `bsub` to submit jobs to the **batch queue**. - Assigns: ✅ **Job name** → `"j1--"` ✅ **Standard output/error logs** 🔹 **Step 5: Monitor Job Progress & Handle Resubmissions** - If `job_resubmission_file` exists, only selected jobs are resubmitted. ========================================================================================================= 📌 **SUMMARY** --------------------------------------------------------------------------------------------------------- ✅ Automates **large-scale TM-align** pairwise structure comparisons. ✅ Runs **distributed computing** over **1000 nodes**. ✅ Implements **job monitoring & error handling**. ✅ Organizes **alignment results** into structured directories. ✅ Cleans up **temporary calculation files** after completion. ========================================================================================================= """ # Cluster information... ################################################################################## cluster = "triton" queue = "normal" # normal or short # Dependencies... ################################################################################## from os import mkdir, listdir from os.path import exists from shutil import rmtree import subprocess # Save locations... ################################################################################## save_directory = "/projectnb/isomlab/dan/in-and-out/output/alphaFold_uniprot_v4/" # calculation_name = "2023.11.12.7tmps/" # calculation_name = "2025.08.20.7tmps-rhod_mglu/" calculation_name = "2025.09.27.7tmps-foldseek_e0p01/" calculation_save_path = save_directory + calculation_name if calculation_name and calculation_name != "/": # Create (or delete and create anew) the directory for program output... if not exists(calculation_save_path): mkdir(calculation_save_path) else: rmtree(calculation_save_path) mkdir(calculation_save_path) # File locations... ################################################################################## where_query_pdb_files_are_located = "/projectnb/isomlab/dan/in-and-out/input/alphaFold_v4/alphaFoldQueryStructures-Triton/rhod/" # where_target_cif_set_files_are_located = "/scratch/alphafold4/submission_sets/" where_target_cif_set_files_are_located = "/projectnb/isomlab/dan/in-and-out/output/alphaFold_uniprot_v4/2025.09.27.7tmps-foldseek_e0p01_models/submission_sets/" superdark_script = "/home/dgi5/scripts/proteomes-214M-v4/1-job-code/1-query/1-superDark.py" # This file is necessary for partial job resubmissions... # If not a job resubmission, set to None. ################################################################################## job_resubmission_file = "" # "/home/dgi5/scripts/proteomes-214M-v4/0-job-management/6_job_resubmissions.txt" job_resubmission_numbers = [] if job_resubmission_file: f = open(job_resubmission_file, "r") for line in f: job_number = line.strip() job_resubmission_numbers.append(job_number) f.close() # Collect the job submission sets files... ################################################################################## set_file_paths = [] if not job_resubmission_file: # Collect the 1,000 job submission sets files... ################################################################################## job_submission_files = sorted(listdir(where_target_cif_set_files_are_located)) for job_submission_file in job_submission_files: set_file_paths.append(where_target_cif_set_files_are_located + job_submission_file) else: # Collect the specific job resubmission sets files... ################################################################################## job_submission_files = sorted(listdir(where_target_cif_set_files_are_located)) for job_submission_file in job_submission_files: set_file_number = job_submission_file.split("-")[1].split(".txt")[0] if set_file_number in job_resubmission_numbers: set_file_paths.append(where_target_cif_set_files_are_located + job_submission_file) # Submit the jobs... ###################################################################################################### # Comment out if doing a full calculation # Condition responding to reviewers # Limits analysis analysis to ~5% of 2.14M, so ~10M job_numbers = {n: n for n in range(1, 31)} for set_file_path in set_file_paths: # Create save locations... ################################################################################################## global_job_number = set_file_path.split("set-")[1].split(".")[0] # Comment out if doing a full calculation if int(global_job_number) not in job_numbers: continue query_label = "job-" + global_job_number logResultsDirectory = calculation_save_path + query_label + "/" # Create (or delete and create anew) the directory for program output... ################################################################################################## if not exists(logResultsDirectory): mkdir(logResultsDirectory) else: rmtree(logResultsDirectory) mkdir(logResultsDirectory) # Set the target paths... ################################################################################################## target_paths = [] f = open(set_file_path, 'r') for line in f: target_paths.append(line.strip()) f.close() # Set the query paths... ################################################################################################## query_paths = [] file_names = sorted(listdir(where_query_pdb_files_are_located)) for file_name in file_names: # if "-ssco.pdb" in file_name: # source_pdb_file_name = file_name # query_paths.append(where_query_pdb_files_are_located + source_pdb_file_name) if ".pdb" in file_name: if file_name.endswith("-ssco.pdb"): continue if file_name.endswith("-s.pdb"): continue if file_name.endswith("-h.pdb"): continue source_pdb_file_name = file_name query_paths.append(where_query_pdb_files_are_located + source_pdb_file_name) # Set the combined paths... ################################################################################################## combined_paths = [] for query_path in query_paths: for target_path in target_paths: combined_paths.append(target_path + "*" + query_path) # Write the job submission file. Each line contains two models to align... ################################################################################################## job_number = 1 f_job_path = logResultsDirectory + "job-"+str(job_number)+"-tmalign-submission.txt" f_job = open(f_job_path, "w") for combined_path in combined_paths: f_job.write(combined_path + "\n") f_job.close() # Create log and error file for the job submission file... ################################################################################################## log_directory = logResultsDirectory + "job-"+str(job_number)+"-tmalign.log" error_directory = logResultsDirectory + "job-"+str(job_number)+"-tmalign-error.log" # Submit the job... ################################################################################################## print("submission-" + str(global_job_number) + "-" + str(job_number)) subprocess.run(["bsub", "-q", queue, "-J", "j1-" + str(global_job_number) + "-" + str(job_number), "-o", log_directory, "-e", error_directory, "python", superdark_script, "-a", f_job_path, "-b", cluster, "-c", str(job_number), "-d", logResultsDirectory])