You need to clarify the type of error encountered. If the software fails to build a pharmacophore model, it might be due to overly rigid settings. Adjusting the number and types of features may resolve the issue.

If your pharmacophore modeling job is failing at the final step, there could be several underlying causes that need investigation. First, examine the software's log files for critical errors such as memory allocation failures (e.g., "OutOfMemoryException"), license expiration notices, or missing dependencies, which are common culprits in tools like LigandScout, Phase, or MOE. Memory issues can often be resolved by increasing allocation through parameters like Java's -Xmx flag, while license problems may require renewal or reinstallation.

Input data quality is another key factor—corrupted or improperly formatted files (PDB, SDF) can cause crashes during alignment or feature extraction. Ensure your protein and ligand structures are properly preprocessed: protonated, minimized, and standardized using tools like OpenBabel. If the error suggests missing pharmacophore features, manually inspect intermediate results to verify ligand alignments and feature maps, adjusting tolerance parameters for distances or angles if necessary.

Hardware limitations frequently disrupt final-stage computations. Monitor system resources (CPU, RAM, disk space) during the run, as large datasets or complex systems may exhaust available memory or storage. Solutions include simplifying the system (reducing ligand count, using coarser grids), running smaller batches, or switching to high-performance computing resources.

Algorithmic failures may arise from insufficient feature consensus among ligands or poorly defined binding sites. In such cases, curate your ligand set to ensure shared pharmacophoric features while maintaining diversity, or explicitly define the binding site using a reference ligand. Parallelization bugs can also cause crashes; try disabling multithreading (e.g., setting n_jobs=1) or ensuring thread-safe file operations.

For systematic debugging, rerun the job with verbose logging, test a minimal case (single protein-ligand pair), and validate the software with tutorial datasets (e.g., PDB 1HCL). If problems persist, consider alternative tools like pharmit or seek community support on platforms like CCL or ResearchGate, providing detailed error logs for targeted assistance. Let me know the specific software and error messages for further troubleshooting.

@Abdulhamid Dehghani

@Hojjat Rezaiezadeh

Thank you so much for the answer and valuable information.

I have another question.

I have prepared the ligands, and now I have different ligands for each chemicals . Would you please tell me how can I chose the best ligand?( My object is developing pharmacophore)