Modeling Pulsed Evolution and Time-Independent Variation Improves the Confidence Level of Ancestral and Hidden State...

Area of Science:

• Evolutionary Biology
• Phylogenetics
• Computational Biology

Background:

• Ancestral state reconstruction is crucial for understanding trait evolution and predicting hidden states.
• Current methods often suffer from high prediction uncertainty due to model misspecification, particularly the constant-rate Brownian motion (BM) model.
• The BM model's assumptions of homoscedasticity and normality are frequently violated in empirical data, leading to unreliable confidence intervals.

Purpose of the Study:

• To develop a novel method for ancestral and hidden state prediction that accurately models trait evolution.
• To improve the reliability of confidence estimates in trait evolution predictions.
• To address the computational expense of existing advanced methods.

Main Methods:

• Developed RasperGade (Reconstructing Ancestral State under Pulsed Evolution in R by Gaussian Decomposition), a new method utilizing the Levy process.
• The Levy process explicitly models gradual evolution, pulsed evolution, and time-independent variation.
• Empirical data from mammalian body size and bacterial genome size were used for validation.

Main Results:

• RasperGade demonstrated superior performance compared to BayesTraits and StableTraits in providing reliable confidence estimates for trait evolution.
• The method is orders of magnitude faster than existing alternatives.
• Empirical data showed that the BM model's residual Z-scores are neither homoscedastic nor normal, confirming model misspecification issues.

Conclusions:

• Accurate assessment of prediction uncertainty is critical for meaningful interpretation in trait evolution studies.
• RasperGade offers a computationally efficient and reliable approach for ancestral and hidden state predictions.
• Rate variation in trait evolution must be assessed, and the quality of confidence estimates examined for continuous traits.