Quantum AI for Medical Research.
Drug discovery, clinical trial optimisation, genomic variant prioritisation, and imaging similarity search — powered by quantum-inspired algorithms and a unified AI-native backend. No quantum hardware. No external services. Callable by AI agents via MCP.
One backend
The entire research stack. In one process.
A typical translational research platform stitches together a vector DB, a graph DB, a workflow engine, a quantum cloud, and a RAG framework. Purple8 replaces all of them.
| Capability needed | Traditional stack | Purple8 |
|---|---|---|
| Compound & trial knowledge graph | Neo4j / Amazon Neptune | Purple8 Hyper Graph |
| Literature & EHR retrieval (RAG) | Pinecone + LangChain | Purple8 RAG — hybrid graph + vector retrieval |
| Molecular ground-state estimation | IBM Quantum / PySCF + cloud | Purple8 Quantum — VQE, CPU-only, in-process |
| Cohort & trial scheduling | Custom ILP solver + Gurobi | Purple8 Quantum — scheduling & annealing |
| Imaging similarity search | Cosine vector search only | Purple8 Quantum — fidelity-based state similarity |
| IRB / regulatory workflow | Jira + manual email chains | Purple8 Journey Engine — SLA + approval gates |
| Audit trail | Separate logging service | Purple8 Hyper Graph — immutable audit trail |
| AI agent interface | Custom API + LangGraph | Purple8 MCP Server — 82 tools, agent-native |
Use cases
Six research workflows, one backend
Drug discovery — binding affinity
Encode a candidate molecule's interaction sites as a Hamiltonian. Loop quantum.vqe_step until the variational energy converges. Compare ground-state energies across candidate compounds stored as graph nodes.
Clinical trial cohort assignment
Assign patients to treatment arms under eligibility criteria, stratification requirements, site capacity, and dosage group balance. QUBO + quantum.scheduling solves this in seconds — on the same CPU as the rest of Purple8.
Medical imaging similarity
Extract radiomics features from CT/MRI scans. Encode them as quantum amplitude states via quantum.encode_state. Retrieve the k most similar prior cases by quantum fidelity — not just cosine distance.
Genomic variant prioritisation
Score variant-gene-disease associations under a sequencing budget as a binary optimisation problem. Structurally equivalent to portfolio optimisation: maximise total information value, respect total cost constraints.
IRB & regulatory workflow
Model the IRB review → Phase I → Phase II → NDA submission pipeline as a Purple8 Journey Engine workflow. SLA monitoring flags overdue submissions; approval gates require PI sign-off before advancing to the next phase.
Research knowledge graph + RAG
Ingest published papers, clinical trial registries, and EHR data. Build a graph of compounds, targets, diseases, and trials. Purple8 RAG retrieves relevant evidence before every optimisation run — grounding agent decisions in the literature.
quantum.* namespace
Quantum MCP tools for medical research
AI agents call these tools directly in natural language. No code written by a researcher.
quantum.vqe_stepVariational Quantum Eigensolver — iterates toward molecular ground-state energy; use for binding affinity estimation
quantum.simulateMPS tensor network simulation — models correlated electron systems and spin chains
quantum.encode_stateStore radiomics or molecular feature vectors as quantum amplitude states, HNSW-indexed by fidelity
quantum.compareFidelity, trace distance, and entanglement entropy between two stored quantum states
quantum.measureSimulate Born-rule measurement of a stored quantum state
quantum.annealGeneric QUBO solver — encode cohort assignment, variant prioritisation, or resource allocation
quantum.schedulingClinical trial scheduling with eligibility constraints, site capacity, and dosage group balance
quantum.optimizeAuto-selects the best solver based on problem size and constraint density
The full stack
Graph, RAG, and workflow — same backend
Quantum tools don't run in isolation. Every optimisation is grounded in a knowledge graph, retrieved from literature, and tracked through a regulatory workflow.
rag.hybrid_queryPurple8 RAG — retrieve relevant literature, clinical notes, and prior trial results before running optimisation
data.ingest_filePurple8 DocIntel — ingest FHIR records, genomic VCF files, lab reports, or research PDFs into the knowledge graph
journey.definePurple8 Journey Engine — model drug approval pipelines or IRB review processes with SLA enforcement and approval gates
journey.resolve_hitlPurple8 Journey Engine — gate molecular screening results behind a principal investigator approval step
Agent-native workflow
Drug screening in 6 agent steps
No human writes application code. The AI agent is the developer.
- 01
Ingest the knowledge base
Purple8 DocIntel loads compound libraries, PubMed abstracts, FHIR records, and prior trial results into the knowledge graph.
- 02
Retrieve relevant context
Purple8 RAG retrieves known binding partners, contraindications, and mechanism-of-action papers for the target protein.
- 03
Encode molecular state
Purple8 Quantum stores the candidate molecule's amplitude vector as a graph node, indexed by quantum fidelity for similarity search.
- 04
Estimate binding affinity via VQE
Loop the Variational Quantum Eigensolver until the variational energy converges — estimating the ground-state energy of the molecular Hamiltonian.
- 05
Compare against candidate library
Purple8 Quantum computes fidelity and trace distance between the new compound and all previously screened candidates in the graph.
- 06
Advance to PI review
Purple8 Journey Engine moves the compound to the Principal Investigator approval gate. Sign-off unblocks the next phase — fully audited, immutable trail.
FAQ
Common questions
How does VQE help in drug discovery?
The Variational Quantum Eigensolver (VQE) iteratively minimises the energy of a Hamiltonian representing a molecular system. By finding the ground-state configuration, it estimates binding affinity and conformational stability — key properties in lead compound selection. Purple8's quantum.vqe_step MCP tool runs this in pure numpy on CPU, iterating toward convergence without quantum hardware.
Can Purple8 optimise clinical trial patient assignment?
Yes. Patient-to-trial assignment is a combinatorial optimisation problem with binary decision variables (enrolled or not), eligibility constraints, dosage group balancing, site capacity, and stratification requirements. This maps directly to QUBO, solved via quantum.scheduling or quantum.anneal. An AI agent can retrieve the patient graph, formulate the QUBO, and call the solver — all through MCP.
How is quantum fidelity useful for medical imaging?
Radiomics feature vectors extracted from CT or MRI scans can be encoded as quantum amplitude states via quantum.encode_state and stored in Purple8's HNSW index. Similarity search then uses quantum fidelity rather than cosine distance — capturing correlations between feature components that standard vector similarity misses. quantum.compare returns fidelity, trace distance, and entanglement entropy between any two stored states.
Does this require quantum hardware or a cloud quantum service?
No. All quantum-inspired tools in Purple8 run on ordinary CPUs using pure numpy — no QPU, no IBM Quantum, no AWS Braket, no external subscription. For production-scale quantum chemistry beyond what VQE covers, the right pattern is to use Purple8 as the knowledge graph, RAG, and workflow backbone, and invoke an external chemistry solver from within a JourneyEngine stage hook.
Ready to accelerate your research?
Developer edition is free. All quantum.*, rag.*, journey.*, and graph.* tools are included. No hardware, no cloud subscription, no special setup.