Universal Computational Intelligence for Life Sciences & Solid-State Materials: Plants & Crops, Animals, Human Therapeutics, Microbial Biocatalysis & Nanotechnology.
Real-time 3D biophysical simulation, 4-Key AI vs Experiment QC Grid, PoseBusters audit, solid-state crystallography, 24+ discovery catalog, and multi-LLM AI Co-Pilot in Polymathix OS v1.0.3.
-00f59b?style=for-the-badge&logo=youtube&logoColor=black){kind=icon}
Polymathix OS is an open-source, AI-orchestrated scientific intelligence operating system designed to accelerate discoveries across all living kingdoms and inorganic solid-state materials:
- 🌾 Agricultural & Crop Sciences: Protein thermal adaptation, drought resilience, Rubisco engineering, natural plant defense metabolites.
- 🔬 Solid-State Materials Science & Nanotech: Cubic Perovskites (
$\text{SrTiO}_3$ ), 2D Graphene monolayers, Carbon Nanotubes ($(5,5)\text{ SWCNT}$ ), Anatase$\text{TiO}_2$ photocatalysts, and Porous MOF-5 lattices. - 🧪 Human Health & De-Novo Drug Design: Small-molecule bioisostere generation, Lipinski Rule-of-Five screening, active pocket docking.
- 🦠 Microbial & Environmental Biocatalysis: Catalytic triad optimization (Ser-His-Asp), plastic-degrading enzymes (PETase), carbon-capture metalloenzymes.
- 🐾 Veterinary & Animal Health: Zoonotic disease targets, species-specific pharmacokinetics, non-toxic veterinary therapeutics.
Rather than acting as a static dashboard, Polymathix introduces a closed-loop AI Orchestrator following the Plan ➔ Execute ➔ Verify methodology.
RESEARCHER
│
Natural Language
│
▼
┌─────────────────────┐
│ AI ORCHESTRATOR │
└──────────┬──────────┘
│
┌───────────────┼───────────────┐
▼ ▼ ▼
PLAN EXECUTE VERIFY
│ │ │
▼ ▼ ▼
Protocol MCP/API Analysis
Manager Router Engine
│ │ │
└───────────────┼───────────────┘
▼
EXTERNAL SCIENTIFIC ENGINES
│
┌──────────────┬─────┴────────┬──────────────┐
▼ ▼ ▼ ▼
RDKit OpenMM PyMOL/RCSB Kernel PCA
│ │ │ │
└──────────────┼──────────────┴──────────────┘
▼
DATA PIPELINE
│
┌──────────────┼──────────────┐
▼ ▼ ▼
Real-Time Three.js 3D Automated
Terminal Logs Viewport Dossier
- Intent-Driven Protocol Selection: Type natural research prompts (e.g., "Synthesize 5-fluoro-aspirin", "Run Crop Proteome & Thermal Adaptation", "Load SrTiO3 Perovskite").
- Multi-LLM 3rd-Party Gateway: Plug-and-play attachment in Settings for Claude 3.5 Sonnet, OpenAI ChatGPT/GPT-4o, xAI Grok-2, Moonshot Kimi, Zhipu GLM-4, Google Gemini 1.5 Pro, and local Ollama instances (100% offline, zero-cloud privacy), with autonomous deterministic fallback.
- Real-Time Structure Mutation: Live conversational chat triggers instant RDKit bioisostere derivation, SMILES conversion, MMFF94 force-field geometry optimization, and direct 3D coordinate injection into the viewport.
- Calibrated Protocols: Pre-calibrated protocols for membrane dynamics, de-novo pharmacophores, non-linear PCA trajectory unfolding, and binding pocket contact mapping.
-
Solid-State Materials Science & Inorganic Crystallography:
-
Cubic Perovskite (
$\text{SrTiO}_3$ ): Cubic space group$Pm\bar{3}m$ (221),$a = 3.905\text{ \AA}$ , 12-fold / 6-fold coordination. -
2D Graphene Sheet: Honeycomb
$sp^2$ Dirac semimetal monolayer lattice. -
Carbon Nanotubes (
$(5,5)\text{ SWCNT}$ ): Chiral armchair 1D ballistic conductors. -
Anatase
$\text{TiO}_2$ : Tetragonal photocatalyst coordination polyhedra. -
Metal-Organic Framework (
$\text{MOF-5}$ ): Nanoporous$\text{Zn}_4\text{O}(\text{BDC})_3$ cage architecture. - Automatic computation of lattice parameters, density, coordination numbers, and pore diameters.
-
Cubic Perovskite (
-
3D Discovery Library Catalog (24+ Targets):
- Curated benchmark systems across Human Therapeutics, Crop Science, Solid-State Materials, Microbial Biocatalysis, and Veterinary Health.
- 1-click "Load 3D Viewport" and "⚡ Make Novel Variant" live derivatization.
-
RDKit Cheminformatics: Automated bioisostere fragmentation, Lipinski Rule-of-Five profiling (MW
$\le$ 500, LogP$\le$ 5, HBD$\le$ 5, HBA$\le$ 10, TPSA$\le$ 140$\text{\AA}^2$ , Rotatable Bonds$\le$ 10), and 2D chemical structure synthesis. - Deep Non-Linear Trajectory PCA: Overcomes linear PCA limitations by deploying RBF Kernel PCA to unroll high-dimensional conformational manifolds.
-
OpenMM Self-Healing Physics: Autonomous PDB topology verification, hydrogen addition at pH 7.4, TIP3P solvation, and potential energy minimization with convergence tracking down to
$-146,240\text{ kJ/mol}$ . -
Live RCSB PDB Ingestion: Direct fetching and atomic parsing of any entry from the Protein Data Bank (e.g.,
1UBQ,4HHB,6LU7,1CRN).
-
Full Pack of 12 3D Scientific Representations: Instant switching between:
-
Ribbon: PyMOL/ChimeraX continuous secondary-structure spline (purple
$\alpha$ -helices, golden$\beta$ -sheets, cyan loops) with C-$\alpha$ nodal beads. -
Cartoon: Classic structural cartoon with cylindrical helical ribbons and directional
$\beta$ -sheet arrow plates. -
Surface: Solvent-accessible surface area (SASA) analytical Conolly boundary (
$1.4\text{ \AA}$ probe). -
Electrostatic: Poisson-Boltzmann continuum Coulombic potential (acidic red
$\to$ neutral white$\to$ basic blue). -
Hydrophobic: Kyte-Doolittle lipophilicity potential mapping (amber lipophilic
$\to$ ice-blue polar). -
Glass: PBR optical transmission dielectric envelope (
$IOR = 1.35$ , transmission$0.88$ , clearcoat$1.0$ ). - Mesh: Triangulated geometric wireframe polygon volume contour.
-
Ball & Stick: Dual-radius valence bonds (
$0.16\text{ \AA}$ ) with CPK atom joints. -
Licorice: Uniform covalent valence sticks (
$0.24\text{ \AA}$ ) across the structure. - Spacefill: Full van der Waals CPK contact spheres (Carbon, Nitrogen, Oxygen, Sulfur, Hydrogen).
-
B-Factor / pLDDT: AlphaFold / crystallographic confidence rainbow spectrum (rigid blue
$\to$ flexible red). - Backbone Trace: Minimalist C-$\alpha$ peptide trajectory spline with glowing nodal markers.
-
Ribbon: PyMOL/ChimeraX continuous secondary-structure spline (purple
- 5 Layer Overlays: Independent toggle controls for S-S Disulfide Bridges, TIP3P Explicit Water Shell, POPC Membrane Bilayer, Ligand Active Pocket, and Binding Cavity Mesh.
-
4 Biophysical Simulation Modes: Real-time Langevin Thermal Dynamics (
$300\text{ K}$ fluctuations with live temperature slider), Normal Mode Analysis (NMA) domain breathing, Induced-Fit Ligand Docking with dynamic contact distance telemetry, and Thermal Denaturation ($350\text{ K}$ ). -
Active Site Contact Profiling: Euclidean residue-ligand distance calculations with live interaction classification (Hydrogen bond, Hydrophobic, Electrostatic,
$\pi$ -$\pi$ stacking). - Multi-Format 3D Exporters: Export models to OBJ (Blender), STL (3D printing), GLB/GLTF (Web & XR), and PLY (CAD/Point Clouds).
-
The "Four Keys in One Grid" Architecture:
-
Key 1 (Fit to Experiment): Kabsch superimposition computing Backbone RMSD, Cα Pocket RMSD (
$0.84\text{ \AA}$ ), and Heavy-Atom Ligand RMSD ($1.12\text{ \AA}$ ). -
Key 2 (Model Confidence): Per-residue pLDDT spectrum highlighting highly confident secondary structures vs flexible peripheral loops (
$83.2$ mean pocket pLDDT). -
Key 3 (Residue Deviation
$\Delta r_i$ ): Euclidean spatial deviation vectors discriminating rigid binding pocket anchors (97 preserved anchors) from flexible hinge breathing (max$2.48\text{ \AA}$ at loop D112). -
Key 4 (Model-Estimated Pairwise Error [PAE]): 2D Predicted Aligned Error heatmap matrix evaluating domain-domain relative orientation confidence (
$2.15\text{ \AA}$ intra-pocket PAE).
-
Key 1 (Fit to Experiment): Kabsch superimposition computing Backbone RMSD, Cα Pocket RMSD (
-
PoseBusters 12-Point Physical Plausibility Audit:
Automated compliance verification checking steric clash ratios (
$0.00$ ), covalent bond lengths ($1.12\sigma$ ), bond angle valence strain ($1.48\sigma$ ), aromatic planarity ($0.03\text{ \AA}$ ), chiral center retention ($100%$ ), and force-field internal strain ($\Delta E = 1.84\text{ kcal/mol} < 5.0\text{ kcal/mol}$ ). -
Flagship Benchmark: BCL-2 Complexed with Navitoclax (ABT-263):
Directly addresses the fundamental scientific question: "When AI and experiment disagree on a binding site, where do you look first?" with the 4-step biophysical triage sequence:
-
Experimental Electron Density & Crystallographic B-Factors (flexible loops exhibit higher thermal displacement
$B > 50\text{ \AA}^2$ ; loop breathing is often conformational sampling rather than an AI error). - PoseBusters Physical Plausibility & Internal Strain ($\Delta E$) (eliminating steric clashes and high-energy unphysical conformers).
- Cross-Domain PAE & Inter-Chain Confidence (confirming domain-domain hinge orientation).
- Protein-Ligand Interaction Fingerprints (PLIF) (confirming key pharmacophoric anchors such as Asp103 H-bonding and hydrophobic pocket insertion).
-
Experimental Electron Density & Crystallographic B-Factors (flexible loops exhibit higher thermal displacement
- 3D Viewer: Full orbit controls, focus, camera reset, snapshot download.
- Sequence Viewer: Primary amino acid sequence chips with residue numbering and secondary structure color coding.
- Trajectory Timeline: Dynamic frame scrubber with RMSD and energy stability tracking.
- Properties Table: Molecular weight, atom count, formula, chains, resolution, isoelectric point (pI), and crystal lattice parameters.
- Analysis: Comparative plots for latent space manifold projections and minimization trajectories.
- Python 3.10 or higher
- Conda or
pip
git clone https://github.com/sohel3shaikh/Polymathix.git
cd Polymathix# Using conda (Recommended for OpenMM and RDKit)
conda create -n polymathix python=3.10 -y
conda activate polymathix
conda install -c conda-forge rdkit openmm pdbfixer -y
pip install -r requirements.txtpython -m streamlit run app.pyOpen your browser at http://localhost:8501.
Run the full end-to-end automated verification suite to test all scientific engines:
python scripts/test_all.pyExpected output:
============================================================
POLYMATHIX OS — COMPREHENSIVE END-TO-END VERIFICATION
============================================================
[TEST] 1. Polymathix Orchestrator Initialization & PDB Parsing... --> PASSED
[TEST] 2. De-Novo Drug Synthesis Engine (RDKit Backend)... --> PASSED
[TEST] 3. Deep Trajectory PCA & Manifold Learning Engine... --> PASSED
[TEST] 4. Self-Healing Energy Minimization & OBJ Exporter... --> PASSED
[TEST] 5. Quick Action Handlers & Session Lifecycle... --> PASSED
[TEST] 6. Agent Controls (Pause, Stop, Run)... --> PASSED
[TEST] 7. Tool Connector Router & Toggle Logic... --> PASSED
[TEST] 8. Frontend Scientific Assets Synchronization... --> PASSED
[TEST] 9. Frontend Index HTML & Three.js 3D Engine Integrity... --> PASSED
[TEST] 10. Streamlit Master Application Compilation... --> PASSED
[TEST] 11. D-Glucose Active Pocket Docking & 3D Interacting... --> PASSED
[TEST] 12. Universal Small Molecule 3D Engine (Aspirin, etc.)... --> PASSED
[TEST] 13. 100% Progress Convergence & Protocol Finalization... --> PASSED
[TEST] 14. Solid-State Materials Science & Crystallography... --> PASSED
[TEST] 15. 3D Discovery Library Catalog (24+) & Real-Time AI... --> PASSED
[TEST] 16. Cross-Domain Physical Invariants & Correctness... --> PASSED
[TEST] 17. Cryptographic Run Provenance & Replay Engine... --> PASSED
[TEST] 18. 4-Key AI vs Experiment QC Grid & PoseBusters Audit... --> PASSED
============================================================
VERIFICATION SUMMARY: 18 PASSED, 0 FAILED
============================================================
Polymathix/
├── app.py # Master Streamlit entry point
├── polymathix/ # Core Python scientific intelligence backend
│ ├── orchestrator.py # AI Orchestrator (Plan -> Execute -> Verify)
│ ├── core/ # Multiscale biophysics & verification
│ │ ├── ai_qc.py # 4-Key AI vs Experiment QC Grid & PoseBusters Audit
│ │ ├── invariants.py # Domain-Agnostic Physical Invariants & Correctness Verifier
│ │ ├── provenance.py # Cryptographic Scientific Provenance & Replay Engine
│ │ └── hybrid_dynamics.py # QM/MM boundary dynamics
│ └── modules/ # Domain-specific scientific tools
│ ├── ai_engine/ # Multi-LLM API Gateway (Claude, ChatGPT, Grok, Kimi, GLM)
│ ├── material_science/ # Solid-State Crystallography (SrTiO3, Graphene, SWCNT, MOF-5)
│ ├── drug_hunting/ # RDKit de-novo synthesis & Lipinski screening
│ ├── quick_pca/ # Deep Kernel PCA non-linear manifold learning
│ ├── data_resilience/ # Potential energy graphing engine
│ └── visual_factory/ # PDB to OBJ 3D mesh exporter
├── polymathix_ui/ # Bidirectional Streamlit Custom Component
│ ├── __init__.py # Python-to-JavaScript component bridge
│ └── frontend/ # Glassmorphic 3D dashboard
│ ├── index.html # Three.js WebGL dashboard & layout
│ └── assets/ # Live scientific plots, models, renders & animated demo
├── scripts/ # Automated testing, video generation & release scripts
│ ├── test_all.py # End-to-end test suite (18 tests)
│ ├── generate_linkedin_video.py # 48-second HD showcase video & GIF builder
│ └── publish_github_release.py # GitHub Releases API publishing automation
├── 1CRN.pdb # Benchmark reference (Crambin plant seed protein)
├── requirements.txt # Python package requirements
└── README.md # Comprehensive scientific documentation
We welcome contributions from structural biologists, crystallographers, materials scientists, bioinformaticians, agronomists, veterinary researchers, and machine learning engineers!
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/NewDiscoveryEngine) - Commit your Changes (
git commit -m 'Add Crop Drought Resilience Module') - Push to the Branch (
git push origin feature/NewDiscoveryEngine) - Open a Pull Request
Distributed under the MIT License. Free for academic, industrial, and open-source scientific research.
@software{polymathix2026,
author = {Polymathix Open-Source Contributors},
title = {Polymathix OS: Autonomous AI for Multi-Kingdom Scientific Discovery},
year = {2026},
publisher = {GitHub},
url = {https://github.com/sohel3shaikh/Polymathix}
}