Descripción
- Plant genetic resources preserved in genebanks have the potential to harbor agronomically important traits that can be used to overcome current and future environmental challenges. However, despite their importance, these resources remain mostly underutilized worldwide. During the last years, the development of genome sequencing technologies has opened new avenues in the discovery of genes relevant for diverse agronomic traits in multiple crop species. Genebanks and the scientific community have generated massive amounts of phenotypic, genotypic, climatic and geospatial data related to germplasm accessions. Currently, the challenge is to efficiently use these high-dimensional data sets to uncover the true potential of plant accessions for breeding purposes. Despite artificial intelligence (AI) tools may outperform competing methods in extracting hidden information from high-dimensional data, their use in the study of genebank accessions is still in its infancy. This situation might be due to the fact that these tools are relatively new, and applications in the field of plant genetic resources are still incipient. To address this problem, we propose to use AI tools to comprehensively analyze phenotypic, genomic and environmental data associated to plant germplasm accessions held at genebanks to reach an in-depth understanding of their genetic diversity, uncover useful germplasm in scenarios of climate change, and promote their utilization. As a case study, we propose to demonstrate the use of AI tools for Phaseolus bean accessions held at genebanks in Colombia, a major agro-biodiversity hotspot. Specifically, we propose to use AI for three main purposes: (1) to identify phenotypic and agronomic traits that are highly informative for the classification and prioritization of germplasm accessions, (2) to identify genetic variants related to agronomic traits and environmental variables of adaptive significance in climate change scenarios and, (3) to generate a free software and graphical interface to help researchers in the analysis and visualization of multidimensional data associated to genebank accessions. Researchers from CGIAR’s International Center for Tropical Agriculture (CIAT), the National University of Colombia, and the Agricultural Research Corporation (AGROSAVIA), the institution that manages the National System of Germplasm Banks of Colombia, have independently produced in the last 30 years vast amounts of morpho-agronomic, genomic, climatic and geospatial data to understand the genetic structure of common bean (Phaseolus vulgaris L.), Lima bean (Phaseolus lunatus L.) and allied species, and the association of genetic variants to relevant adaptive traits. In this proposal, the three institutions will carry out a collaborative effort to strengthen their research capabilities and perform comprehensive analyses of high-dimensional data associated to Phaseolus germplasm accessions deposited at the genebanks of CIAT and AGROSAVIA. The data that these three institutions have already produced consist of: (1) genotyping-by-sequencing data (GBS) of a collection of 582 wild and domesticated accessions of common bean and Lima bean that cover the geographic range and genepools within these species, (2) the assembly of a reference genome for a Mesoamerican landrace of Lima bean, and (3) passport and morpho-agronomic data associated to 37,938 accessions of 44 Phaseolus taxa from 112 countries held at both genebanks. Additional (4) resequencing public data of 60 wild individuals and 100 landraces will be included as genetic controls. These datasets are already organized and labeled to be used as training data. The timing for the three key milestones of the project is as follows: (1) Classification and prioritization of Phaseolus accessions on the basis of phenotypic and genomic data analyzed with machine learning and deep learning tools (month 4), (2) identification of genetic variants associated to agronomical variables and envi