Laboratory workflows

Laboratory work crosses from scientific advice into physical execution. The workflow therefore requires an explicit promotion from advisory planning to a reviewed executable plan, followed by a separate observed-outcome record.

Plan follow-up

Begin with a governed program and evidence bundle. Build an advisory assay plan that maps open evidence needs to assay recommendations and wet-lab actions. Review the objective, blocking posture, rationale, sample kind, and unresolved evidence for every recommendation.

An advisory plan is never an operator instruction. It carries plan_kind="advisory" and executable=false so that serialization cannot erase the boundary.

Design and batch work

Translate accepted advice into an experimental design with explicit contrasts, biological and technical replication, randomization, blocking, fractionation, multiplex roles, controls, and carryover risk.

plan_experiment_batches() groups assays by blocking status and assay family, orders declared dependencies, attaches review gates, and records sample requirements. Reject dependency cycles and missing prerequisites before scheduling.

flowchart TD
    advice["advisory assay plan"]
    design["reviewed experimental design"]
    batches["dependency-ordered batches"]
    readiness["operational readiness"]
    executable["executable instructions"]
    outcome["observed outcome"]
    reconcile["reconciliation and feedback"]
    advice --> design --> batches --> readiness
    readiness -->|ready| executable --> outcome --> reconcile
    readiness -->|not ready| refusal["execution refusal"]

Establish readiness

Assess materials, reagents, controls, provenance, evidence sufficiency, staffing, instrument availability, review backlog, and assay-specific risk. Capacity and scheduling reports should expose delayed or infeasible batches rather than silently reordering scientific priority.

When prerequisites are insufficient, emit a LabExecutionRefusal with stable reason codes and required corrections. A refusal is the correct outcome when execution would be irresponsible.

Create the handoff

Select one reviewed batch and build executable instructions. Each instruction names its assay, batch, sample kind, objective, blocking posture, and preflight checks. Attach protocol version, preparation metadata, instrument method, controls, failure caveats, material reservations, and the authority boundary.

Wrap the typed payload in a canonical artifact envelope and verify its digest before export. A LIMS payload is an operational view; preserve the canonical plan when field mapping loses nested rationale or policy.

Capture outcomes

Record raw replicate values, summary statistic, dispersion, unit, QC state, normalization, censoring, detection limit, batch-effect notes, and interpretation confidence. Classify failed outcomes as technical, biological, reproducibility, material, or interpretation failures rather than reducing every result to pass/fail.

Apply the declared rerun policy and evaluate evidence-promotion readiness. A result blocked by QC remains an outcome but does not become trusted knowledge evidence.

Reconcile and continue

Compare requested and observed assays, plan deviations, expected and observed states, resource use, and unresolved evidence. Emit a reconciliation report, operator follow-through actions, intelligence feedback, and a next-cycle plan when justified.

Do not replace the original recommendation, plan, or expected outcome. Link new records to them so later reviewers can determine what was known and authorized at each boundary.

See laboratory planning and outcome contracts and laboratory artifact contracts.