Exponentially Modified Peak Functions in Biomedical Sciences and Related Disciplines

Area of Science:

• Biomedical data analysis
• Statistical modeling in biology

Background:

• Many biological processes involve variable times between states, often modeled by probability distributions.
• The exponentially modified Gaussian (EMG) distribution is common in chromatography and psychophysiology, and fits biological timing data like cell division and gene expression better than simpler models.
• A limitation of EMG is its potential to produce negative values, which is physically unrealistic in many biological contexts.

Purpose of the Study:

• To review and compare the applicability of the exponentially modified Gaussian (EMG) and exponentially modified gamma-distribution (EMGD) for modeling biological timing data.
• To highlight the advantages of EMGD over EMG when dealing with data where non-negativity is a physical requirement.

Main Methods:

• Review of existing literature on the application of EMG and EMGD in various biological and physiological contexts.
• Discussion of the mathematical properties and parameter interpretability of both distributions.
• Comparison of fitting performance and physical interpretability for biological data.

Main Results:

• Both EMG and EMGD provide superior fits to biological timing data compared to lognormal, gamma, and Wald distributions.
• EMG's unlimited Gaussian component can lead to unrealistic negative predictions, especially with high variance.
• EMGD, incorporating a non-negative gamma distribution, offers a more physically meaningful alternative, particularly when the EMG Gaussian component's coefficient of variance exceeds approximately 0.4.

Conclusions:

• EMGD is a preferable model to EMG for biological timing data when non-negativity is essential and the Gaussian component's variance is substantial.
• The choice between EMG and EMGD depends on the specific data characteristics and the physical constraints of the biological system being modeled.