Reference Architecture
Hide
Layered architecture showing how Research Hub sits at the front as the unified consumption app for research and discovery users while orchestrating AI, analytics, data, and source systems. Business and scientific purpose: this view supports faster, evidence-based drug discovery decisions across research, operations, and governance teams. Reference Architecture is connected to linked entities, metrics, and workflow actions in this workspace. Use these panels to review evidence, identify risks, and decide the next best scientific action. The view uses deterministic demo data shaped to mirror realistic drug discovery operations.
Users / Personas Layer
Layer 1
Plan studies, run experiments, and review evidence for progression decisions.
Execute assays, manage protocols, and capture ELN records with traceability.
Analyze omics data, run compute pipelines, and validate statistical outputs.
Run SAR, docking, and molecular profiling for hit-to-lead optimization.
Assess PK/PD, toxicology, and animal study evidence for translational readiness.
Connect preclinical signals to biomarker and candidate advancement decisions.
Track milestones, dependencies, and resource alignment across portfolios.
Prioritize programs using risk/benefit signals and decision dossiers.
Monitor policy adherence, CAPA, audit readiness, and data governance controls.
Prepare submission evidence, trace lineage, and verify controlled exports.
Own metadata quality, lineage continuity, and master data consistency.
Manage AI safety controls, agent behavior, and operational reliability.
Application / Consumption Layer
Layer 2
Unified front-end operating system for all research and discovery users and use cases.
Research Hub Capabilities Layer
Layer 3
Studies, experiments, protocols, ELN, assays, samples, and instruments.
Data hub, datasets, lineage, metadata, knowledge graph, literature, and patents.
Notebooks, pipelines, runs, batch orchestration, and storage governance.
Compound library, SAR, docking, protein structure, and ADMET workflows.
Imaging operations, analytical visualization, and phenotype insight views.
Preclinical, PK/PD, toxicology, animal studies, and biomarker readiness.
Portfolio prioritization, milestones, risks, Go/No-Go, and reports.
Threads, decisions, exports, integrations, and cross-team traceability.
Assistant controls, audit log, consent/privacy, and settings governance.
Domain agents and copilot experiences with controlled execution and auditability.
Generative AI Layer
Layer 4
Guided research help with governed interactions.
Specialized agents for protocol, chemistry, translational, and reporting tasks.
Reusable, role-specific templates for scientific and business use cases.
Safety checks, confidence framing, approval gates, and auditable run logs.
Drafts experiment protocols, acceptance criteria, and safety checkpoints.
Proposes controls, replicates, and sample plans for robust study design.
Summarizes SAR trends, suggests analog directions, and flags outliers.
Explains docking pose confidence and potential binding-site concerns.
Highlights developability and safety risks from multi-assay signatures.
Detects assay anomalies and recommends rerun or QC investigation paths.
Finds missing metadata, schema drift, and suspicious lineage breaks.
Narrates derivation chains and downstream impact for decisions.
Builds evidence summaries across publications and competing hypotheses.
Summarizes claim clusters, overlap risk, and freedom-to-operate signals.
Generates scenario comparisons for milestones, risk, and value tradeoffs.
Checks action requests against SOP, consent/privacy, and policy constraints.
Coordinates multi-step task chains across data, compute, and reporting.
Drafts structured evidence narratives for export and submission workflows.
Analytics & Compute Layer
Layer 5
Jupyter/RStudio/SAS style interactive analysis environments.
Airflow, Nextflow, Argo, Snakemake, and CWL execution layers.
Slurm clusters, Kubernetes, serverless jobs, and GPU pools.
Docking, MD simulation, FEP, and conformer generation engines.
NGS alignment, variant calling, RNA-seq, proteomics, and metabolomics.
Feature pipelines, model training/registry, and inference services.
PK/PD modeling, dose-response analysis, and translational statistics.
Run telemetry, quality gates, anomaly alerts, and failure triage.
Automated report generation, dashboarding, and decision briefs.
Data & Knowledge Layer
Layer 6
Raw, curated, and serving zones for assay, omics, and image data.
Analytics-ready tables, semantic models, and governed marts.
ELN attachments, instrument files, microscopy images, and reports.
Near-real-time instrument, workflow, and integration event ingestion.
Canonical IDs for compounds, targets, studies, samples, and sites.
Business/technical metadata, data owners, glossary, and stewardship.
End-to-end lineage from source systems to decisions and exports.
Completeness, conformance, consistency, and timeliness controls.
Entity-relationship network linking biology, chemistry, and outcomes.
Unified retrieval across notebook text, datasets, reports, and evidence.
Literature, patents, clinical registries, and public database feeds.
Retention, access controls, consent/privacy tags, and compliance labels.
Systems & Source Layer
Layer 7
Benchling, eLabFTW, Signals Notebook, Dotmatics, and custom ELNs.
Sample lifecycle, chain-of-custody, and assay request workflows.
Scientific data and chromatography systems (e.g., CDS/SDMS repositories).
Lab execution, SOP controls, deviations, CAPA, and quality events.
High-throughput screening campaign systems and hit triage outputs.
Liquid handlers, robotic stations, scheduling, and execution logs.
Illumina, PacBio, ONT and associated run files/metrics.
LC-MS/GC-MS vendor outputs and run metadata streams.
Cytometer files, gating outputs, and panel configuration records.
High-content imaging systems and image-derived feature sets.
In vivo study scheduling, cohort records, and endpoint capture.
Specimen repositories, freezer maps, and custody records.
Registration, inventory, plate maps, and structure repositories.
Registration, similarity search, and molecular property tooling.
Reagent procurement, vendor master, spend, and lot intake.
CTMS, EDC, ePRO, and safety systems for translational linkage.
Adverse event and safety signal systems for risk monitoring.
Submission publishing, document control, and correspondence tracking.
SSO, IAM, PAM, and role directory synchronization.
Email, chat, document repositories, and meeting transcripts.
Partner data exchanges, milestones, and deliverable packages.
ChEMBL, PubChem, UniProt, PDB, Ensembl, GEO, and similar sources.
USPTO/EPO and publication providers for competitive intelligence.
Object/file stores used for raw data and generated artifacts.
Jupyter, MLflow, Airflow, DataHub, OpenMetadata, Superset, Keycloak.