ABSTRACT
Stony coral tissue loss disease (SCTLD) remains a substantial threat to coral reef diversity already threatened by global climate change. Restoration efforts and effective treatment of SCTLD requires an in-depth understanding of its pathogenesis in the coral holobiont as well as mechanisms of disease resistance. Here, we present a supervised machine learning framework to describe SCTLD progression in a major reef-building coral, Montastraea cavernosa, and its dominant algal endosymbiont, Cladocopium goreaui. Utilizing support vector machine recursive feature elimination (SVM-RFE) in conjunction with differential expression analysis, we identify a subset of biologically relevant genes that exhibit the highest classification performance across three types of coral tissues collected from a natural reef environment: apparently healthy tissue on an apparently healthy colony, apparently healthy tissue on a SCTLD-affected colony, and lesion tissue on a SCTLD-affected colony. By analyzing gene expression signatures associated with these tissue health states in both the coral host and its algal endosymbiont (family Symbiodiniaceae), we describe key processes involved in SCTLD resistance and disease progression within the coral holobiont. Our findings further support evidence that SCTLD causes dysbiosis between the coral host and its Symbiodinaiceae and additionally describes the metabolic and immune shifts that occur as the holobiont transitions from a healthy to a diseased state. This supervised machine learning framework offers a novel approach to accurately assess the health states of endangered coral species and brings us closer to developing effective solutions for disease monitoring and intervention.
AUTHOR SUMMARY Coral reefs are under increasing threat due to climate change, with rising ocean temperatures and disease outbreaks accelerating reef degradation. Stony coral tissue loss disease (SCTLD) has been particularly destructive, leading to widespread coral mortality across Florida’s Coral Reef and the wider Caribbean since its emergence in 2014. While the cause of SCTLD remains unknown, the rapid decline in coral reef health highlights the urgent need for innovative approaches to understanding threats to coral health. In this study, we applied a supervised machine learning approach, previously used in cancer research, to identify key genes associated with SCTLD progression in the coral Montastraea cavernosa and its symbiotic algae, which the coral relies on to meet its nutritional requirements. By analyzing gene expression patterns across tissues representing different health states, we find that SCTLD affects the metabolic interactions between the coral and their symbionts and causes shifts in coral immune signaling pathways, even in tissue on a SCTLD-affected colony that appears to be healthy. This study presents a novel framework for applying supervised machine learning in coral gene expression research and could lead to new methods for monitoring coral health and combatting SCTLD.
Competing Interest Statement
The authors have declared no competing interest.
