CDD Vault curve fits need assay context before compound ranking

Keep observations separate from the fitted curve

CDD's official page supports a narrow product claim: CDD Vault can organize experimental results and entities, preserve experiment lineage, compare assay data, and generate or fit dose-response curves. The measured observations and the fitted object must remain separately addressable. Preserve every concentration-response pair, replicate, control, missing value, transformation, normalization step, and source import before storing parameters such as potency, efficacy, slope, or other model output.

The curve record should identify the fit model and version, equation or named family, constraints, weighting, initial values, convergence state, treatment of censored values, minimum point rules, replicate aggregation, uncertainty outputs, residual diagnostics, analyst action, and time computed. A graph image, headline parameter, or successful fit status cannot explain which observations drove the result or whether another defensible model would change the scientific interpretation.

Bind every result to assay and entity context

CDD presents chemical or biological registration, activity data, assays, protocols, runtime details, and visualization as connected capabilities. The buyer should test whether that connection remains exact at decision time. Retain the registered entity and salt or construct, lot or batch, sample and plate well, concentration and unit, dilution series, reagent lots, cell or biological system, instrument or readout, protocol and version, run identifier, operator, date, environmental or incubation conditions, and the controls used to qualify the run.

Identity corrections and protocol changes need dated lineage. A structure update, reassigned sample, unit conversion, plate-map repair, control-limit change, or revised normalization should identify every curve and comparison affected without overwriting the prior state. If data arrive by API, parser, spreadsheet, instrument connection, or manual entry, preserve the original payload, mapping version, rejected rows, duplicate rule, and reconciliation totals so clean visualizations do not conceal an incomplete or mis-keyed experiment.

Make exclusions and comparison sets reviewable

Compound ranking adds another decision layer. Record the population considered, assay and endpoint chosen, replicate rule, comparator and reference controls, transformation, aggregation across runs, threshold, weighting across endpoints, treatment of inconclusive or toxic observations, and owner of the ranking question. A single best-looking curve should not silently replace a declared comparison set, and a structure-activity visualization should show which experiment versions and filters produced it.

Every exclusion needs a reason linked to the original observation, applicable procedure or scientific judgment, reviewer, date, and effect on the fit and rank. Test a failed control, partial curve, nonmonotonic response, unit mismatch, duplicate entity, plate effect, saturated readout, outlier removal, model that does not converge, and a rerun that conflicts with the first experiment. The system should present uncertainty and disagreement for accountable review rather than manufacture scientific validity from a smooth line.

Read CDD Vault within its evidence boundary

CDD's official record establishes current provider positioning for experimental-data management, entity registration, assay context, lineage, curve fitting, visualization, APIs, automation, and related informatics capabilities. It does not establish a customer's configuration, source completeness, entity correctness, protocol suitability, control acceptance, model fitness, reproducibility, interpretation, candidate quality, regulatory use, or scientific conclusion. Provider claims about precision, rigor, and decision support require representative buyer testing.

Lab Systems Index reviewed the official record on September 2, 2026. No exact dated material development after the September 1 publication cutoff was established, so this is durable scientific-data analysis rather than a current-intelligence event. A buyer test should reconstruct one ranked compound from registered entity and assay request through raw observations, controls, units, normalization, model and exclusions, curve version, comparison set, independent review, decision, and later reproduction.

Enterprise buyer test

Translate this change into the exact population, record type, workflow stage, decision owner, effective date, and evidence that could be affected. Ask current or prospective providers to demonstrate the named workflow with representative data and an exception—not a polished feature tour. Record what official documentation establishes, what a provider states, what the team observes, and what remains unresolved.

A defensible review also identifies the dependency outside the product. Authority interpretation, policy configuration, data quality, integrations, human judgment, approval rights, release governance, training, and retained evidence may remain customer or service responsibilities. The evaluation should preserve those boundaries instead of treating a technology claim as the complete operating model.

What we will watch next

Lab Systems Index will watch the named source and affected market records for later evidence that changes status, scope, availability, implementation timing, workflow consequence, or the limits of the initial report. A later announcement does not silently overwrite this dated account; the change ledger preserves the sequence.