Format: ORAL

Irene Cobo-Simn1, Nicols Lobato2, Julio Ramrez-Guerrero2, Lorenzo Federico Manjarrez1, Nuria de Mara1, Miriam Lpez-Hinojosa1, Mara Dolores Vlez1, Paula Ramos1, Jos Antonio Mancha1, Alberto Pizarro3, Jos Antonio Cabezas1, Mara ngeles Guevara1, Carmen Daz-Sala3, Mara Teresa Cervera1

1 Institute of Forest Science - National Institute for Agricultural and Food Research and Technology, Spanish National Research Council (ICIFOR-INIA, CSIC), Madrid, Spain. 2 University of Seville, Seville, Spain. 3 University of Alcalá (UAH), Alcalá de Henares, Spain.

Knowledge about the molecular and functional mechanisms that control the adaptation of trees to drought is key in the current context of climate change, since recurrent drought is one of the main climatic factors threatening forests, particularly in the Mediterranean region. However, this knowledge is scarce, especially in conifers, due to their huge and complex genomes. Pinus pinaster is one of the conifers with the greatest socioeconomic (for wood and resin production) and ecological importance in southwestern Europe, given its adaptive capacity. Recent advances in sequencing technologies enable the generation of large amounts of omics data, introducing a new era of big data in biology. Machine learning (ML), a branch of Artificial Intelligence (AI), offers promising approaches to recognize fine-grained patterns and relationships in these large and heterogeneous datasets, helping bridge the gap between our capability to produce and analyze plant molecular data. Here, we explore the potential of machine learning in predicting drought tolerance-related molecular biomarkers in maritime pine (P. pinaster). To this end, we used transcriptomic data from grafted rootstocks of genotypes with contrasting response to drought (tolerant and sensitive). Machine learning algorithms were used to analyze the expressed genes on these grafted rootstocks under controlled drought conditions to obtain accurate predictions of their tolerance/sensitiveness to drought. The methodology includes data selection and preprocessing, feature selection, model development, and evaluation. The obtained drought-tolerance molecular biomarkers can be used to guide the design of breeding and management programs, such as marker assisted selection and breeding, which will help improve the conservation, health and productivity of this conifer in the long term. They can be also used in future drought-response research (e.g. comparative genomic studies, functional transfer across species) which will shed light on trees molecular adaptive potential to current climate change.