Plant breeding for resistance to insectsHow sensor technology and artificial intelligence contribute to
sustainable crop production
Lucas P.J.J. Noldus
Sjoukje Heimovaara
Marcel Dicke
Maarten A. Jongsma
LUCAS NOLDUS
▪ M.Sc. Leiden University, The Netherlands (biology)
▪ Ph.D. Wageningen University, The Netherlands (entomology)
▪ Founder and CEO, Noldus Information Technology
▪ Research Associate, Wageningen University & Research
▪ Chairman, ICT for Brain, Body & Behavior Foundation
▪ Secretary, Man-Machine Interaction Platform
▪ Secretary for International Affairs, Netherlands Academy of Technology
and Innovation
SPEAKER BIOGRAPHY
Noldus Information Technology
THE GLOBAL CHALLENGE
AGRICULTURE IN THE
NETHERLANDS
September 2017
The Netherlands: 2nd agricultural export country in the world
21
BIOTIC AND ABIOTIC
FACTORS IMPACTING
CROPS
▪ Bacteria
▪ Fungi
▪ Viruses
▪ Nematodes
▪ Insects
▪ Extreme temperatures
▪ Salt
▪ Drought
Thrips
Apids
Aphids, thrips, bugs, leafhoppers, whiteflies: sucking insects
PLANT DAMAGE CAUSED
BY SUCKING INSECTS
▪ Tissue damage
▪ Virus transmission
▪ Cosmetic damage
27 April, 2018
CONVENTIONAL APPROACH:
INSECTICIDES
CROP RROTECTION
ALTERNATIVE: BREEDING FOR
RESISTANCE TO INSECTS
▪ Phenotypic screening of many plant accessions
▪ Find resistance genes/markers against pest insects
▪ Understand molecular basis of resistance mechanism
▪ Breed resistant variety
CROP RROTECTION
SCREENING PLANTS
FOR RESISTANCE TO
INSECTS
Conventional screening methods▪ Assess feeding damage▪ Count offspring and survival▪ Measure development timeAll done by visual rating and scoring
Drawbacks▪ Labor-intensive, costly▪ Time-consuming▪ Imprecise▪ Subjective
Needed: automated
solution!
VIDEO TRACKING
Video image of insect
Digitized object with center of body mass (x,y) and track
Movement track before and after smoothing
Movement classes
FROM MOVEMENT TO
FEEDING BEHAVIOR
Video tracking system measures movement
▪ Distance
▪ Velocity
▪ Turning angle, meander
What we need to know
▪ Probing behaviour
▪ Feeding behaviour
How to assess feeding behaviour from movement parameters?
VIDEO TRACKING
AUTOMATIC DETECTION
OF FEEDING BEHAVIOR
VIDEO TRACKING
VALIDATION OF
AUTOMATED BEHAVIOR
RECOGNITION
Top view →EthoVision video tracking software
Side view →The Observer behavioral scoring software
VIDEO TRACKING
Validation of automated behavior recognition
Short probes
< 3 min
Long probes
> 25 min
Nick Sloff, Plant Physiology (2012)
PROBE DURATION
INDICATES FEEDING
Automated video tracking: validation
0
High-throughput screening
High-throughput screening
Up to 200 assays running in parallel (using one video camera)
Lactuca sativa – Nasonovia ribisnigri Arabidopsis thaliana – Myzus persicae
Probe duration on resistant vs. susceptible plants
What is the effect of leaf discs?
Electrical Penetration Graph
Resistance attenuated, but detectable (n=25)
EPG Video tracking
Plant material Intact plants Leaf discs
Minimum observation duration 4-8h 4-8h
Number of replicates (identification rate ± 80%) 3-5 20-25
Sensitivity of plant effects high low
Maximum sample size per set up ± 8 ± 100
Preparation time per sample ± 5-10 min ± 2 min
Assessment of electrical patterns/video images 15 min Automated
PHENOTYPING
METHODS COMPARED
FROM NO-CHOICE TO
TWO-CHOICE ASSAYS
Prison plate (a)
Choice plate (c)
Blocking plate (b)
6 mm
Cover plate (d)
e
6mm
6mm
1mm
1mm
Blocking plate (b)
Prison plate (a)
Choice plate (c)6 mm
Cover plate (d)
e
6mm
6mm
1mm
1mm
▪ No-choice: insect is placed on
leaf disc
▪ Two-choice: insect can choose
between two leaves
Release compartment closed
Release compartment open
Apparatus for high-throughput 2-choice tests
Top view
Cross section
Arena design in video tracking software
55 dual-choice assays, each consisting of two leaf discs (green) and a release compartment (white)
Automated video tracking
55 dual-choice assays, each consisting of two leaf discs (green) and a release compartment (white)
RESULTS OF 2-CHOICE
ASSAYS
▪ Video tracking shows clear
preference for susceptible
variety
▪ After 8 hours instead of 6
days!
FROM LEAF DISCS TO
INTACT LEAVES
Prototype setups with intact leaves
Heated top plate and LED lightsArena template foam
From leaf discs to intact leaves
Four leaves with >60 arenas
SOFTWARE TOOLS
EthoVision
▪ Video tracking system
▪ Experiment automation
▪ From video to movement tracks
▪ Up to 200 arenas running in parallel
EthoAnalysis
▪ Imports track files from EthoVision
▪ Data selection and filtering (time, zone, etc.)
▪ Computation of numerous end-points
▪ Statistical analysis
VIDEO TRACKING
COMPARED WITH OTHER
PHENOTYPING METHODS
Video tracking → high throughput
▪ 1 plate/hour filled
▪ >200 arenas/plate
▪ 20 arenas/genotype
▪ >10 genotypes/plate
▪ ~50 genotypes/person/day
90% reduction in time and costs!
VALIDATION WITH
COMMERCIAL INSECT-
PLANT COMBINATIONS
Insect species
▪ Aphids
▪ Thrips
▪ Whiteflies
Vegetables
▪ Pepper
▪ Tomato
▪ Water melon
▪ White cabbage
▪ Lettuce
▪ Bitter gourd
Ornamental plants
▪ Chrysanthemum
▪ Lily
NEXT STEPS
Software engineering
▪ Protocols and tools for integration of
resistance screening with genetics
▪ Using more advanced AI techniques
(e.g. deep learning)
Hardware engineering
▪ From prototype to commercial
product
CONCLUSIONS
▪ Video technology, computer vision and pattern recognition enable
automated high-throughput screening of plants for resistance to
sucking insects
▪ Novel method can lead to 90% reduction in time and costs of
screening
▪ Genetic improvement of commercial crops will reduce dependency
on chemical pesticides
▪ Result: more sustainable crop production
ACKNOWLEDGEMENTS
Wageningen University & Research
▪ Karen Kloth
▪ Manus Thoen
▪ Harro Bouwmeester
▪ Johannes Kruisselbrink
▪ Leo Poleij
▪ Gerrie Wiegers
Noldus Information Technology
▪ Wil van Dommelen
▪ Olga Krips
Plant breeding companies
▪ Royal van Zanten
▪ Syngenta
▪ Bayer
▪ East-West Seed
MORE INFORMATION?
Noldus Information Technology BV
Wageningen, The Netherlands
Email: [email protected]
Web: www.noldus.com
Noldus Information Technology Latin America SpA
Santiago, Chile
Email: [email protected]
Web: www.noldus.com/es