Estimation of the neutrality index.

Area of Science:

• Evolutionary biology
• Population genetics
• Molecular evolution

Background:

• The McDonald-Kreitman (MK) test compares nonsilent and silent substitutions/polymorphisms to infer selection.
• The neutrality index (NI) is commonly used to quantify deviation from neutral evolution.
• Averaging NI values across genes is a standard approach to summarize species-level selection patterns.

Purpose of the Study:

• To identify and address statistical bias in the McDonald-Kreitman test when averaging neutrality index values.
• To propose alternative statistical methods for accurately assessing selection patterns, especially with sparse data.

Main Methods:

• Analysis of simulated and real biological data using the McDonald-Kreitman test.
• Application of a variant of the Cochran-Mantel-Haenszel procedure for estimating weighted average odds ratios.
• Development and evaluation of a new statistic, the direction of selection (DoS).

Main Results:

• The standard method of averaging neutrality index values across genes introduces significant statistical bias.
• This bias can misrepresent species as undergoing purifying selection when they exhibit adaptive evolution, and vice versa.
• A weighted average odds ratio and the direction of selection (DoS) statistic effectively mitigate this bias.

Conclusions:

• The conventional averaging of MK test results is statistically flawed and can lead to erroneous conclusions about evolutionary selection.
• A weighted average odds ratio provides a less biased estimate of selection patterns.
• The proposed direction of selection (DoS) statistic offers a robust alternative for studying selection, particularly when data are limited.