Simulation data for an estimation of the maximum theoretical value and confidence interval for the correlation...

Area of Science:

• Computational chemistry and cheminformatics
• Pharmacokinetics and drug delivery research

Background:

• Assessing the reliability of predictive models is crucial, especially with variable experimental data.
• Understanding the correlation coefficient's (r) confidence interval and maximum theoretical value aids in evaluating model performance.
• This work builds upon previous research using molecular dynamics for in silico skin permeability screening.

Purpose of the Study:

• To present data from numerical simulations quantifying the impact of data uncertainty on predictive model reliability.
• To illustrate how increasing data uncertainty affects the maximum theoretical correlation coefficient (r).
• To determine the confidence interval of r using the Fisher r→Z transform.

Main Methods:

• Conducted purposely designed numerical simulations.
• Analyzed the effect of increasing data uncertainty on the maximum theoretical correlation coefficient (r).
• Employed the Fisher r→Z transform to calculate the confidence interval of r.

Main Results:

• Demonstrated that increased data uncertainty significantly worsens the maximum theoretical correlation coefficient (r).
• Provided simulation data illustrating this relationship.
• Quantified the impact of uncertainty on model reliability metrics.

Conclusions:

• Data uncertainty is a critical factor that degrades the reliability of predictive models.
• The presented data and methodology aid in estimating the trustworthiness of in silico screening models.
• Accurate assessment of correlation coefficient confidence intervals is essential for robust computational predictions.