Accelerating Climate Resilient Plant Breeding by Applying Next-Generation Artificial Intelligence. Issue 11 (November 2019)

• Record Type: Journal Article
• Authors: Harfouche, Antoine L.; Jacobson, Daniel A.; Kainer, David; Romero, Jonathon C.; Harfouche, Antoine H.; Scarascia Mugnozza, Giuseppe; Moshelion, Menachem; Tuskan, Gerald A.; Keurentjes, Joost J.B.; Altman, Arie
• Is Part Of: Trends in biotechnology. Volume 37:Issue 11(2019)
• Page Start: 1217
• Page End: 1235
• Publication Date: 2019-11
• Subjects: next-generation artificial intelligence -- explainable AI -- field phenomics -- genomics -- augmented breeding -- smart farming; Biotechnology -- Periodicals; Biochemical engineering -- Periodicals; Genetic engineering -- Periodicals; Industrial microbiology -- Periodicals; 660.605
• DOI: 10.1016/j.tibtech.2019.05.007
• Languages: English
• ISSNs: 0167-7799

Abstract

Abstract : Breeding crops for high yield and superior adaptability to new and variable climates is imperative to ensure continued food security, biomass production, and ecosystem services. Advances in genomics and phenomics are delivering insights into the complex biological mechanisms that underlie plant functions in response to environmental perturbations. However, linking genotype to phenotype remains a huge challenge and is hampering the optimal application of high-throughput genomics and phenomics to advanced breeding. Critical to success is the need to assimilate large amounts of data into biologically meaningful interpretations. Here, we present the current state of genomics and field phenomics, explore emerging approaches and challenges for multiomics big data integration by means of next-generation (Next-Gen) artificial intelligence (AI), and propose a workable path to improvement. Highlights: The integration of genomics and phenomics will speed the development of climate resilient crops; however, these omics technologies are generating large, heterogeneous, and complex data much faster than currently can be analyzed. First-generation AI is being used in surveying and classifying omics data; however, it is designed to solve well-defined tasks of single-omics datasets that do not require integration of data across multiple modalities. Next-generation AI can change the dynamics of how experiments are planned, thus enabling better data integration, analysis, and interpretation. There is a critical need to develop means by which to open the black boxes prevalent in many current AI approaches so that they can be interpreted meaningfully from a complex biological perspective. AI decisions and outputs can be explained by breeders and researchers via human–computer interaction.