openpkflow.dissolution¶
Dissolution similarity analysis: f1/f2, bootstrap, model fitting, CSV loading, study-level comparison.
Public API¶
| Symbol | Type | Description |
|---|---|---|
f1(reference, test) |
function | Difference factor |
f2(reference, test, method) |
function | Similarity factor; method="regulatory" applies FDA 85% rule |
max_deviation(reference, test) |
function | Maximum absolute deviation (FDA/SUPAC-IR 1995) |
msd_vessels(reference, test, similarity_limit_pct=10, confidence_level=0.90) |
function | Vessel-level Mahalanobis distance with pooled covariance and the Tsong et al. (1996) 90% Hotelling region vs. similarity-limit decision |
MSDVesselResult |
dataclass | .msd, .ci_lower, .ci_upper, .similarity_limit, .is_similar, .summary() |
msd(reference, test) |
function | Mean-profile MSD. Degenerate (MSD squared is always n - 1), so it emits a UserWarning; use msd_vessels |
MSDResult |
dataclass | Mean-profile MSD result: .msd, .msd_squared, .is_similar, .summary() |
regulatory_cutoff(reference, test) |
function | Number of leading timepoints kept by the FDA 85% rule |
model_dependent_comparison(ref_t, ref_Q, tst_t, tst_Q, model) |
function | Compare fitted dissolution model parameters via 90% CI |
ModelComparisonResult |
dataclass | Model comparison result: .ratio_pct, .ci_lo, .ci_hi, .is_similar |
bootstrap_f2(reference, test, ..., f2_method="all_points") |
function | Bootstrap CI for f2; f2_method="regulatory" resamples on the 85%-rule timepoints |
BootstrapF2Result |
dataclass | Bootstrap result container |
DissolutionStudy |
class | Multi-batch study; .compare(), .fit_models(), .bootstrap_compare() |
ComparisonResult |
dataclass | f1/f2 comparison result; .summary(), .plot(), .report() |
fit_dissolution_models(time_points, observed, label) |
function | Low-level model fitting |
DissolutionFitResults |
dataclass | Ranked model fits; .best, .summary(), .plot(), .report() |
ModelFit |
dataclass | Single model fit: params, R2, AIC, AICc, BIC, .predict() |
load_dissolution_csv(path, config) |
function | CSV loader with validation |
DissolutionCSVConfig |
dataclass | CSV column config |
get_formulation_means(df, label) |
function | Extract mean profile for a formulation |
validate_dissolution_dataframe(df, config) |
function | Validate and normalize in-memory vessel data |
DissolutionWorkbenchConfig |
dataclass | Exact workbench configuration |
DissolutionWorkbenchResult |
dataclass | Complete result with .report() and .audit_bundle() |
run_dissolution_workbench(data, config) |
function | Run the complete vessel-level workflow |
run_dissolution_workbench_csv(path, config) |
function | Run the complete workflow from CSV |
Models available in fit_dissolution_models¶
zero_order, first_order, higuchi, korsmeyer_peppas, weibull
Ranked by AICc (small-sample-corrected). Pass models=["weibull", "first_order"] to fit a subset.
Report formats¶
.report("out.html") — HTML
.report("out.md") — Markdown
.report("out.pdf") — PDF (requires openpkflow[reports])
.report("out.docx") — Word (requires openpkflow[reports])
Workbench reports support HTML, PDF, and DOCX. The workbench audit ZIP contains normalized input, exact configuration, serialized results, HTML report, and a SHA-256 manifest.