Entos Qcore

Official Resources

• Homepage: https://entos.ai/
• Documentation: https://docs.entos.ai/ (if available) / Academic papers
• Source Repository: Proprietary / Free for Academic
• License: Free Academic License / Commercial

Overview

Entos Qcore is a modern quantum chemistry software package developed by Entos, Inc. It emphasizes the integration of physics-based methods with machine learning (ML), specifically featuring Molecular Orbital Based Machine Learning (MOB-ML). It is designed for efficiency and ease of use, providing standard DFT and wavefunction methods alongside innovative ML-accelerated approaches.

Scientific domain: Quantum chemistry, Machine Learning, DFT
Target user community: Academic researchers, pharmaceutical industry, materials design

Theoretical Methods

• Density Functional Theory (DFT)
• Hartree-Fock (HF)
• Molecular Orbital Based Machine Learning (MOB-ML)
• Semi-empirical methods (xTB interface)
• Geometry optimization
• Ab initio molecular dynamics
• Composite methods

Capabilities (CRITICAL)

• Fast DFT calculations
• MOB-ML for high-accuracy energies at low cost
• Reaction path finding
• Conformational search
• Geometry optimization
• Vibrational frequency analysis
• Modern C++ architecture
• Python API (EntosQ)

Key Strengths

MOB-ML:

• Machine learning on molecular orbitals
• Transferable accuracy
• Correct physics scaling
• Reduced training data needs
• Accelerates high-level methods

Modern Design:

• Efficient C++ core
• User-friendly input
• Structured output (JSON)
• Clean Python integration
• robust optimizers

Performance:

• Optimized integral engines
• Fast SCF convergence
• Efficient threading
• Scalable algorithms

Inputs & Outputs

• Input formats:
  • Qcore input format (structured)
  • XYZ / PDB coordinates
  • Python API calls
• Output data types:
  • JSON structured output
  • Energies, Gradients
  • Properties
  • Log files

Interfaces & Ecosystem

• Python: EntosQ python library
• OrbNet: Integration with Entos AI tools
• Visualization: Standard tools via output formats

Advanced Features

Machine Learning Integration:

• Training MOB-ML models
• Using pre-trained models
• Active learning workflows
• Uncertainty estimation

Geometry Optimization:

• Robust coordinate systems
• Transition state search
• Constrained optimization
• Reaction path following

Performance Characteristics

• Speed: Competitive with modern C++ codes
• Accuracy: DFT and ML-corrected accuracies
• System size: Medium to large molecules (ML accelerated)
• Memory: Efficient handling
• Parallelization: SMP/OpenMP

Computational Cost

• DFT: Standard scaling O(N^3-N^4)
• MOB-ML: Significantly faster than target method (e.g., CCSD(T))
• Training: One-time cost for ML models
• Inference: Very fast

Limitations & Known Constraints

• License: Proprietary (Free Academic)
• Source: Not open source
• Methods: Focused on DFT/ML, less legacy method coverage
• Documentation: Access may depend on license

Comparison with Other Codes

• vs Gaussian/ORCA: Entos emphasizes ML integration (MOB-ML)
• vs Psi4: Entos has proprietary ML features
• vs TeraChem: Both commercial, Entos focuses on ML/CPU, TeraChem on GPU
• Unique strength: MOB-ML for high-accuracy acceleration

Application Areas

Drug Discovery:

• Binding Affinity: High-throughput calculation of accurate protein-ligand binding energies
• Conformational Analysis: Rapid screening of low-energy conformers using ML potentials
• pKa Prediction: Fast and accurate free energy calculations
• Virtual Screening: Filtering large compound libraries with physics-based accuracy

Catalysis & Reaction Engineering:

• Transition States: Efficient location of transition structures for reaction mechanisms
• Barrier Heights: Accurate activation energies using ML-corrected DFT
• Catalyst Screening: High-throughput evaluation of catalyst performance
• Mechanism Exploration: Interactive path searching on ML surfaces

Best Practices

Licensing & Setup:

• Academic Access: Register for the free academic license to access full features
• Installation: Use provided binary packages or containers for immediate deployment
• Environment: Set up creating isolated conda/python environments for EntosQ

Machine Learning Workflows:

• Training Data: Ensure training sets cover the relevant chemical space of your problem
• Validation: Always validate ML predictions against full DFT for a subset of systems
• Active Learning: Use uncertainty queries to iteratively improve models

Calculation Strategy:

• Geometry Opt: Use robust internal coordinates for flexible molecules
• SCF Convergence: Utilize appropriate DIIS or second-order convergence accelerators
• Python API: Leverage EntosQ for complex, multi-step workflows not possible in single inputs

Community and Support

• Entos, Inc. support
• Academic user community
• Caltech/Miller group origins
• Publications and webinars

Verification & Sources

Primary sources:

1. Homepage: https://entos.ai/
2. Miller group publications (Caltech) on MOB-ML
3. Manby et al. publications

Confidence: VERIFIED

• Status: Active commercial/academic software
• Technology: MOB-ML published
• Developer: Entos, Inc.