Biodiversity: New approach for detecting and delimiting species with AI

What is a species? How were species defined in the past and how will they be described in the future? How many species are there on earth? How many are still undiscovered? Can we describe them faster than they become extinct due to climate change or human influences?

From Greek philosophers such as Aristotle to Charles Darwin and Lamarck to the present day, scientists have been grappling with these fundamental questions. Contrary to public perception, however, they have remained largely unresolved to this day.

Modern genomics teaches us today that what we call species can be unfounded entities based on purely morphological, regional descriptions of species from the past. This applies in particular to groups of animals, plants and fungi that are characterized by complex evolutionary processes such as hybridization or asexuality. This is where the challenges of so-called integrative taxonomy, which combines methods of classical taxonomy in a multidisciplinary way, for example with molecular biological methods, external appearance or ecological studies, become clear: more than 30 species concepts, lack of universal characteristics or markers, lack of suitable analysis tools for large data sets and complex evolutionary processes, as well as strongly author-dependent data integration.

To address this, an interdisciplinary and international research team of biologists and computer scientists from Germany, Spain, China and the USA, led by Dr. Kevin Karbstein, Lara Kösters, Dr. Ladislav Hodač, Martin Hofmann, Dr. Jana Wäldchen and Prof. Dr. Patrick Mäder from the Max Planck Institute for Biogeochemistry and the Ilmenau University of Technology have addressed these questions and written an innovative review article that has now been published in the renowned journal Trends in Ecology and Evolution (TREE; Open Access Link: https://www.cell.com/trends/ecology-evolution/fulltext/S0169-5347(23)00296-3).

In this article, the authors present the vision of a modern integrative taxonomy based on a unified species concept in combination with artificial intelligence (deep learning), which combines the discovery of genetic units with the fusion of automatically extracted information such as morphology, physiology, ecology or behavior to discover species as natural units. This process is known as species delimitation. In this way, artificial intelligence can help accelerate the decoding of biodiversity on an unprecedented scale.