Revealing Extinction Risk through Comparative Genomics and Artificial Intelligence

Descripción del proyecto

Genómica e inteligencia artificial para predecir el riesgo de extinción de especies

Comprender por qué se extinguen las especies es una cuestión básica en la biología evolutiva y la conservación de la biodiversidad. La lista roja de especies amenazadas de la Unión Internacional para la Conservación de la Naturaleza utiliza factores como el tamaño poblacional y las amenazas externas para predecir la tasa de disminución y el riesgo de extinción de cada especie. Sin embargo, pasa por alto factores genéticos que pueden aportar información sobre la diversidad genética y las mutaciones perjudiciales que pueden influir en el riesgo de extinción. En el proyecto REVEAL, financiado por las acciones Marie Skłodowska-Curie, se combinarán datos genómicos de aves a gran escala, simulaciones informáticas e inteligencia artificial para descubrir firmas genéticas vinculadas al riesgo de extinción. Los investigadores evaluarán las ventajas y limitaciones de emplear genomas de referencia para predecir el riesgo de extinción y así fundamentar las estrategias de conservación.

Objetivo

Understanding species extinction risk is a crucial goal in Evolutionary Biology and a contemporary societal challenge. The standard
method for assessing extinction risk, the IUCN Red List, focuses on external threats and overlooks genetic factors. Despite some weak
and variable correlations, genetic diversity offers insights beyond the Red List, potentially resulting in an underestimation of
extinction risk. Genomic data, generated rapidly through initiatives like the Bird 10,000 Genomes Project, has the potential to
enhance assessments by revealing historical population demography, genetic diversity, and the genetic load of deleterious
mutations. Recently, there has been excitement about the potential of reference genomes in conservation genomics. However, the
actual potential and limitations of single reference genomes in informing conservation strategies remain unexplored. To address this
gap, I propose a comprehensive approach, REVEAL, which integrates comparative genomics and individual-based simulations within
a robust Artificial Intelligence (AI) framework. The proposal comprises three steps: (i) simulating extinction risk using a broad
parameter space with a dataset of at least 3,825 bird genomes generated by the B10K consortium, (ii) training AI models to recognise
genomic signatures associated with extinction risk, based on the previous simulations, and (iii) evaluating the potential and
limitations of single reference genomes for extinction risk assessment, using the trained AI models. In particular, I will focus on avian
species with diverse geographic ranges and ancestral population sizes in order to compare extinction risk predictions, considering
both spatial dynamics and temporal dynamics. REVEAL seeks to move beyond conventional assessments and improve our
understanding of extinction risk evaluation using genomic data, ultimately enhancing our ability to formulate effective species
recovery strategies.