Patterns from the
Life Science Industry
A clinical study, a medicinal product, a drug batch, a regulatory submission, a safety
case, or a quality event is not a pile of documents.
This page shows why the
usual ways of managing these concepts and corresponding digital artifacts fall short of
the principles on the previous page, and the pattern that fixes them.
Why current approaches break down
A cell membrane rejects what does not fit. A shared spreadsheet cannot.
Biomedical R&D is digitally inefficient because its data do not function as one connected system. Documents and siloed applications keep information, knowledge, decisions, agents, and interfaces as separate components instead of parts of one governed entity. That is not a tooling problem; it is a conceptualization and modeling problem, and it shows up the same way at multiple stages of the value chain.
Trapped in silos
Data sits in separate systems across macro-siloes (research, development, manufacturing, and commercialization) or micro-silos (personal computers, platforms). Nobody has the full picture/understanding.
Inconsistent meaning
The same concepts/entities (e.g. molecule, study, product, or asset) are described with different words, different attributes based on specific domain vocabularies that humans and IT systems are using.
Disconnected history
Facts are cut off from their lineage across the lifecycle; provenance is reconstructed under time pressure instead of carried by design.
Hidden dependencies
Scientific, business, operational, regulatory or other constraints live apart from the facts they govern, so validity rules survive only as tribal knowledge or isolated documents.
Conflicting lenses
Each stakeholder reads the same object through different responsibilities and incentives, and each builds its own slide, report, or dashboard.
Unvalidated knowledge
An exploratory research association is silently reused as a clinical-selection criterion, regulatory assertion, or commercial claim, with no record of the conditions, confidence, or contradicting evidence under which it holds.
Decisions that never learn
A dose change, a batch release, or a program stop is filed as meeting minutes, cut off from the evidence behind it and the outcomes that follow, so neither failures nor successes teach the organization anything.
Think of your business like a living cell
Treat data, information, knowledge, and decisions as parts of one living system, not disconnected files and departmental handoffs.
Biomedical R&D becomes more efficient when the whole lifecycle behaves like a cell: four layers that keep meaning intact from discovery through commercialization.
Canonical data
One governed source of truth for every molecule, study, product, and asset.
Context-specific views
The right expression for researchers, clinicians, manufacturers, regulators, and commercial teams.
Validated knowledge
Models, rules, and evidence folded into something that actually works.
This model preserves meaning across the full lifecycle, connects scientific evidence to regulatory and operational constraints, exposes dependencies, governs AI, retires obsolete knowledge, and keeps every decision traceable to its evidence and consequences.
Miosis proposes the "holon" approach to solve these issues: a holon is a single governed knowledge object that keeps one identity across the whole lifecycle and is composed of four coordinated graphs:
- a scene graph (what is true now)
- a boundary graph (what must hold)
- an event graph (what happened)
- a projection graph (who needs which view), each detailed below
Scene graph
What is currently asserted?
The scene graph is the object's present state: what is true right now for this specific clinical study, medicinal product, or drug batch, kept as one governed record instead of scattered digital artifacts across systems.
Trial X; protocol v3.0; 82 sites; 640 enrolled patients; endpoint Y; current data-cut D; open safety signals; upcoming milestones.
Boundary graph
What is allowed, required, or forbidden?
The boundary graph holds the rules of validity. The scene graph says what is true; the boundary graph says what must be true, as machine-checkable constraints instead of rules buried in multiple document (SOPs, technical requirements, regulatory concerns…).
ICH GCP requirements; inclusion/exclusion criteria; visit-window rules; endpoint-derivation rules; controlled terminology; authority-specific submission constraints.
Event graph
What happened, when, by whom, and under what authorization?
The event graph is the provenance layer: not just current truth, but defensible truth. It behaves like a committed audit ledger across the object's whole lifecycle, not an afterthought reconstructed before an inspection.
Protocol amendment approved; site activated; patient consented; visit completed; lab result received; query resolved; database locked.
Projection graph
Who needs which view, at what resolution?
The projection graph generates role-specific views from the same governed object. Each audience sees the projection it needs: the document, dashboard, or package is computed, not the raw data, not a duplicated and/or modified version of the raw data.
Investigator view; monitoring dashboard; medical-monitor safety view; CSR tables; regulatory submission; inspection package; executive progress view.
One holon, four graphs
The clinical study report, monitoring dashboard, submission dataset, and inspection package are not disconnected artifacts. They are different projections of the same governed clinical-study holon.
Scene: protocol v3.0, 82 sites, 640 patients, endpoint Y, data-cut D.
Boundary: ICH GCP, protocol constraints, visit windows, inclusion/exclusion, endpoint and dataset rules.
Event: amendment approved, site activated, patient consented, visit completed, lab received, database locked.
Projection: protocol synopsis, monitoring dashboard, CSR tables, submission, inspection package, executive view.
Holons turn regulated life-science artifacts from documents into computable, auditable knowledge objects that expose their state, rules, history, and role-specific views.