Trait-mediated speciation and human-driven extinctions in proboscideans revealed by unsupervised Bayesian neural...

Area of Science:

• Evolutionary Biology
• Paleontology
• Computational Biology

Background:

• Species richness dynamics are shaped by life-history traits, paleoenvironment, and biotic interactions.
• Traditional birth-death models often assume simple, monotonic relationships, limiting analysis of complex ecological factors.

Purpose of the Study:

• To introduce a novel Bayesian birth-death model employing unsupervised neural networks.
• To explore multifactorial and nonlinear influences on speciation and extinction rates using fossil data.
• To disentangle the relative importance and interactions of various predictors on diversification.

Main Methods:

• Development of a Bayesian birth-death model integrated with unsupervised neural networks.
• Application of explainable artificial intelligence (XAI) techniques to interpret model outputs.
• Analysis of the proboscidean fossil record to infer speciation and extinction dynamics.

Main Results:

• Speciation rates in proboscideans were significantly influenced by dietary flexibility and major biogeographic events.
• The emergence of modern humans was identified as a key factor escalating extinction rates.
• Regional climate exhibited a comparatively minor impact on recent proboscidean diversity decline.

Conclusions:

• The developed model offers a more nuanced understanding of clade diversification by analyzing complex, nonlinear relationships.
• Human emergence has had a profound, accelerating effect on extinction rates, leading to recent diversity loss.
• Future research can leverage this model to explore intricate evolutionary dynamics across diverse taxa.