Application of high throughput phenomics in lettuce

admin

(Contributed by Li Bo)

       Plant phenomic studies began at the end ofthe 20th century[1,2].  In the past 20 years, scientistshave developed a series of high-throughput, high-precisionphenotypic acquisition tools, including non-invasive imaging, robotics,reflection spectroscopy, machine vision, environmental sensing, andhigh-throughput cell phenotypic screening.  Relying on these devices,it has been possible to record phenotypes at various scales from seeds, roots,leaves, cells, tissues and canopies, individuals and groups, providingcomprehensive phenotypic evidence for plant researches[3].  Phenomic data, combined with thecorresponding genotype data and environmental data, play an increasinglyimportant role in plant breeding[4].  In recent years,combined with machine learning, artificial intelligence and other dataprocessing methods, plant phenomic research is entering a new stage ofdevelopment. Phenomic can not only help breeders understand complex agronomictraits, but also analyze many previously unknown new traits.  At thesame time, it promotes the development of scientific research and application[5,6,7,8].Here we summarizedrecent advances in studies of application of highthroughput phenomics in lettuce.

       Du etal.[9] utilized a "Sensor-to-Plant" greenhousephenotyping platform to periodically capture top-view images of lettuce, anddatasets of over 2000 plants from 500 lettuce varieties were thus captured ateight time points during vegetative growth. This study presented a novel objectdetection-semantic segmentation-phenotyping method based on CNNs to conductnon-invasive and high-throughput phenotyping of the growth and developmentstatus of multiple lettuce varieties. An object detection model was used todetect and identify each pot from the sequence of images with 99.82% accuracy,semantic segmentation model was utilized to segment and identify each lettuceplant with a 97.65% F1 score, and a phenotyping pipeline was utilized toextract a total of 15 static traits (related to geometry and color) of eachlettuce plant. Furthermore, the dynamic traits (growth and accumulation rates)were calculated based on the changing curves of static traits at eight growthpoints. Finally, validated the application of image-based high-throughputphenotyping through geometric measurement and color grading for a wide range oflettuce varieties.

       Han et al.[10] obtained the floral opening time of 236 RILswas scored using time-course image series obtained by drone-based phenotyping.Floral pixels were identified from the images using a support vector machinewith an accuracy >99%. A Bayesian inference method was developed to extractthe peak floral opening time for individual genotypes from the time-stampedimage data. Two independent quantitative trait loci (QTLs; Daily Floral Opening2.1 and qDFO8.1) explaining >30% of the phenotypic variation in floralopening time were discovered. Candidate genes with non-synonymous polymorphismsin coding sequences were identified within the QTLs. This study demonstratesthe power of combining remote sensing, machine learning, Bayesian statistics,and genome-wide marker data for studying the genetics of recalcitrantphenotypes.

Reference:

[1] RibautJM, de Vicente MC, Delannay X. Molecular breeding in developing countries:challenges and perspectives [J]. Curr Opin Plant Biol. 2010,13, 213–218.

[2] TesterM, Langridge P. Breeding technologies to increase crop production in a changingworld [J]. Science. 2010, 327, 818–822.

[3] FinkelE. With 'Phenomics', plant scientists hope to shift breeding into overdrive[J]. Science. 2009, 325, 380–381.

[4] BevanMW, Uauy C, Wulff BB, Zhou J, Krasileva K, Clark MD. Genomic innovation forcrop improvement [J]. Nature. 2017, 543(7645):346-354.

[5] FurbankRT, Tester M. phenomics--technologies to relieve the phenotyping bottleneck[J]. Trends Plant Sci. 2011,16(12):635-44.

[6] TardieuF, Cabrera-Bosquet L, Pridmore T, Bennett M. Plant phenomics, from sensors toknowledge [J]. Curr Biol.2017, 27, R770–R783.

[7] RoitschT, Cabrera-Bosquet L, Fournier A, Ghamkhar K, Jiménez-Berni J, Pinto F, OberES. Review: new sensors and data-driven approaches—a path to next generationphenomics [J]. Plant Sci. 2019, 282, 2–10.

[8] Mir RR,Reynolds M, Pinto F, Khan MA, Bhat MA. High-throughput phenotyping for cropimprovement in the genomics era [J]. Plant Sci. 2019, 282, 60–72.

[9] Du J, Lu X, Fan J, Qin Y, Yang X, Guo X. Image-BasedHigh-Throughput Detection and Phenotype Evaluation Method for Multiple LettuceVarieties[J]. Front. Plant Sci., 2020, 11:1-15.

[10] Han R, Wong AJY, Tang Z, Truco MJ, Lavelle DO, KozikA, Jin Y, Michelmore RW. Drone phenotyping and machine learning enablediscovery of loci regulating daily floral opening in lettuce[J]. Journal ofExperimental Botany, J Exp Bot., 2021, 72(5).