A standalone, dependency-free Python implementation of weighted average
slope (weighted AS or ASw) from observed concentration-time data. It uses
all consecutive observations from t = 0 through observed Tmax, inclusive,
and places progressively greater emphasis on the earlier intervals.
For n observations up to and including Tmax:
slope[i] = (C[i+1] - C[i]) / (t[i+1] - t[i])
weight[i] = (Tmax - t[i]) / Tmax
1 n-1
weighted average slope = ----- × Σ weight[i] × slope[i]
n - 1 i=1
The weight uses the left-hand time of each interval. The denominator is
n - 1, the number of intervals; the weights are not normalized by their
sum. This detail is important because normalizing by sum(weight) would
produce a different metric from the published ASw definition.
The units of weighted AS are concentration/time. Compared with the unweighted AS, it emphasizes early observations, where the concentration-time profile is more strongly associated with the absorption process.
The definition follows:
Karalis VD. On the Interplay between Machine Learning, Population Pharmacokinetics, and Bioequivalence to Introduce Average Slope as a New Measure for Absorption Rate. Applied Sciences. 2023;13(4):2257. https://doi.org/10.3390/app13042257
Python 3.9 or newer is required.
python -m pip install .Run the tests without third-party testing packages:
python -m unittest discover -s tests -vfrom weighted_average_slope import weighted_average_slope
result = weighted_average_slope(
times=[0, 0.5, 1.5, 3, 5],
concentrations=[0, 2, 6, 9, 7],
)
print(result.weighted_average_slope) # 2.7777777777777777
print(result.tmax) # 3.0
print(result.interval_slopes) # (4.0, 4.0, 2.0)
print(result.interval_weights) # (1.0, 0.833..., 0.5)The bundled example contains two subjects:
weighted-average-slope examples/concentration_time.csv --group subjectWrite results to a CSV file:
weighted-average-slope input.csv --group subject --output weighted_as_results.csvUse different column names:
weighted-average-slope input.csv --time TIME --concentration CONC --group IDThe output includes ASw, Cmax, Tmax, the number of points, and semicolon- separated slopes, weights, and weighted slopes so that every calculation can be audited. Input rows may be unordered; the CLI sorts each group by time.
- Time must start at exactly
0, be finite, non-negative, and strictly increasing within a profile. - Concentrations must be finite and non-negative.
- At least one post-dose point is required before or at the peak.
- The first occurrence of a tied observed Cmax defines Tmax by default. Use
--tmax-tie lastto select the last occurrence or--tmax-tie errorto reject ties. - Missing and BLQ values are rejected rather than silently imputed. Apply and document the preprocessing rule specified in the analysis plan first.
- ASw is calculated independently for each profile/subject. Concentrations should not be averaged across subjects before calculation.
This software implements a calculation and is not validated for clinical or regulatory decision-making. Users are responsible for data review, validation, pre-specified handling rules, and compliance with applicable guidance.
src/weighted_average_slope/ library and CLI
tests/ unit and CLI tests
examples/ example concentration-time CSV
CITATION.cff citation metadata
LICENSE MIT license
.github/workflows/ automated tests on GitHub