1 · Set constrained molar composition
All four components sum to exactly 100 mol%. The bounds describe a feasible region, not independent lipid-factor ranges.
Process factors
Process factors are independent of lipid composition and use low, center and high coded levels. Disable a process factor to fit a mixture-only model.
2 · Scientific DOE checklist
- Composition: four mole fractions form a simplex; the generated settings also respect user-specified lower/upper limits.
- Design candidate set: constrained polytope vertices plus interior interpolations; rank-first/space-filling selection. This is not a certified D-optimal design.
- Regression: Scheffé mixture components and pairwise terms with lipid–process interactions. No independent intercept, avoiding collinearity with the 100% constraint.
- Replication: repeated measurement of one batch is not an independent formulation replicate. Design independent batches and track biological unit and time/lot as separate covariates.
- QC mapping: the QC Studio must use the DOE run ID (such as MP001) as its sample ID to transfer matching results.
- Evidence: fitted correlations are hypotheses. Validate on new batches and compare against pre-defined CQAs.
Model structure
Σxᵢ = 1; zₖ ∈ {−1,0,+1}
3 · Randomized LNP formulation worksheet
Generate the design, select the measured response to edit, and enter one value per run. Exporting CSV includes all four response columns; reimport is matched to run IDs and original factor values.
No study generated yet.
4 · Research response model
The model fits the selected measured response with full-rank least squares. It reports coefficients and model-fit descriptors without declaring an optimal lipid formulation.
Enter measured responses, then choose a response and fit the model.
5 · Independent open-source analysis
Use the exported worksheet with R or Python for a full experiment-design audit, residual diagnostics, model selection, prediction uncertainty and prospective confirmation.
NEXUS does not run or bundle third-party R/Python packages. Confirm their current licenses, documentation and supported statistical assumptions independently.