Issuu on Google+

10/19/2011

From Genotype to Phenotype

Ph 1

Gerie van der Heijden (gerie.vanderheijden@wur.nl)

5

Sequencing + Genotyping

Plant Phenotyping at Wageningen UR

Crops

4

2

3

Phenodays, Wageningen October 13, 2011

Genotype……………………………...Phenotype

Environmental variation

Improved crops

Phenotyping at Wageningen UR

Image analysis and automation Vision system for sorting of orchids

Several examples focussing on image analysis      

Robotics and automation of pot plants Time monitoring of Arabidopsis Chlorophyll fluorescence Hyperspectral imaging 3D imaging Analysis of complex plant scenes Contact: Rick vd Zedde

Automatic harvesting of roses

Plantalyser system for pot plants 

Features that can be measured:      

Plant height and width (generative and vegetative) Projected area of leaf and flower (side and top) Number and size of flowers Average colour of leaves and flowers Leaf orientation Shoot density and width Contact: Jochem Hemming

Contact: Rick vd Zedde

1


10/19/2011

Plantalyser  

 

Monitoring Arabidopsis

Virtually for all kind of potplants (max. plant height: 120 cm) Wide range of features are tested and validated for: • Anthurium • Spathiphyllum • Kalanchoë • Curcuma New species will require some tuning New features will require algorithm development and programming

Monitoring Arabidopsis

Hyperspectral imaging

J. Kokorian, G. Polder, J.J.B. Keurentjes, D. Vreugdenhil, M. Olortegui Guzman. An ImageJ based measurement setup for automated phenotyping of plants. ImageJ Conference. 2010

Acquisition of hyperspectral images

Hyperspectral imaging of tomatoes

Tomatoes in different ripeness stadia

2


10/19/2011

Measurement of Lycopene content

Detecting Fusarium in wheat seeds Use of transmission NIR (900-1800 nm)

A. color reflection B. NIR transmission at 1100 nm C. Predicted Fusarium concentration

HPLC measurement G. Polder, G.W.A.M. van der Heijden, H. van der Voet, and I.T. Young. Postharvest Biology and Technology, 34(2):117–129, 2004.

Detecting Fusarium in wheat seeds

Hyperspectral imaging of grassland

PLS with cross validation on seed kernel basis

Predicted Ct

21 --808 --403 --606 --601 --710 --711 --703 --706 --304 --407 --805 --708 --610 --809--807 --810 --603 --302 --802 --211 --307 --412 --602 --605 --611 --812 --104 --505 --401 --702 --508 --202 --509 --310 --308 --212 --507 --102 --301--312 --707

20 19 18 --305 17 --208

--110 --503 --203 --510 --411 --512 --108 --309 --501 --311 --206 --205--511 --409 --210

16 15

--408

14 13

15.5

16

16.5

17

17.5

Object error: RMSEP: 0.79 RMSEP/: 0.04

Pixel error: RMSEP: 0.82 RMSEP/: 0.04

Q2: 0.80

Q2: 0.79

18

18.5

19

19.5

20

20.5

Measured Ct

velocity 0.3-0.5 m/s

G. Polder, G.W.A.M. van der Heijden, C. Waalwijk, and I.T. Young. Seed Science and Technology. 33(3):655–668, 2005.

Predict biomass, DMC and N-content DM yield (ton/ha)

Measured

4

DM content (%)

Physical and chemical measurement

N content (%)

cal

Close sensing

cal

Remote sensing

5

30 25

3

Combining remote and close sensing

4

20 2

3

15 1

2

10

0

5 0

1

2 3 -1 DS opbrengst (ton ha )

4

1

5

10

15

20

Predicted DS gehalte (%)

25

30

1

2

3

4

5

Voorspeld N (%)

A.G.T. Schut, G.W.A.M. van der Heijden, I. Hoving, M.W.J. Stienezen, F.K. van Evert, and J. Meuleman. Agronomy Journal. 2006.

3


10/19/2011

Biomass and nitrogen prediction in the field

CF Transient Imager

Spatial pattern of the predicted biomass (tons/ha) for the grass/clover field

van der Heijden, G.W.A.M., Clevers, J.G.P.W. and Schut, A.G.T. International Journal of Remote Sensing, 28:24, 5485 – 5502. 2007.

Contact: Henk Jalink

Imaging the induction curve of photosynthesis Fv/Fm = (Fmax– F0)/Fmax Fmax

F0

Quantification of stress in leaves

Time sequence cabbage plants After movement correction we can monitor every pixel in time and early discern regions with decreased photosynthesis efficiency

G. Polder, G.W.A.M. van der Heijden, H. Jalink and J. Snel. Computers and electronics in Agriculture, 55:115, 2007.

4


10/19/2011

Salinity stress in potato

High-throughput 3D seedling sorter

Traits:

Tolerant

• Photosynthetic activity Fv/Fm • Distribution of Fv/Fmover plant Control

6D Salt

Find responsible genomic region: QTL

13D Salt

Aim:

Approach: 

Sensitive

Seedling assessment based on human expert knowledge modelling Highly accurate 3D reconstruction using volumetric intersection of 10 cameras simultaneously

Result: 

a high-speed sorting device in cooperation with Flier Systems BV.

Contact: Henk Jalink and Gerard vd Linden

High-throughput 3D-based seedling sorting

Plant Phenotyping in EU project SPICY 

Capacity: 20.000 seedlings/ hour. Processing time: 45ms per seedling Contact: Rick van de Zedde

  

Combine phenotypic data with genotypic data and crop QTL growth models for pepper Partners:      

SPICY: Plant material and genetic map Recombinant inbred lines of pepper Yolo Wonder

CM334

P3 0,0

P4 0,0

F5YC genotyping 297 RIL

7,3

38,6 42,4 44,5 49,3 56,4

65,7 63,4 70,1 76,6 80,9

P2

95,7

85,3 90,6 93,8

0,0

102,0 17,8 18,5 22,2

36,3 40,7

P5 0,0

15,0 19,1

30,8

111,8 130,5 137,9 146,4 153,0

28,2 33,3 35,3 37,2 41,8 51,6 55,8 60,1 62,2 66,9 69,9 72,3 74,3 89,5 97,6 101,9 104,0 106,1 108,5 112,3 114,2 116,7 119,4 123,7 128,7 134,6 137,7 142,6 145,2 148,7

53,6 160,7 62,4 67,1 70,2

91,1 93,6 95,9 101,5 104,6 110,7

172,9

P9 0,0 8,8

13,2

13,7

P10 0,0

7,8

35,2 45,3 47,5

60,4 62,5

55,7

40,0

52,9

34,6

20,6 24,4 28,3 31,2 34,1 36,1 45,1 54,4

180,8 184,9 188,6 192,4 197,8

params

P12 0,0 6,5

23,1

0,0 40,2 46,9

0,0 19,6

20,3

35,3 38,0

110,6 111,9 114,6 115,7 119,2 121,2 124,3 128,3 132,1 133,7 136,2 139,6 144,0 148,7 154,0 157,1 168,5 180,5 185,3 189,5 194,0

28,9

41,5 45,1 46,8

43,9

Build high rig with:

62,0

33,0

90,4 93,4 100,1 107,0 108,3 109,5

54,9

28,8 31,7

76,7

P7 9,0

type

Model

13,2

44,0 55,6

70,7 74,8 86,7

Pheno-

11,4

25,6

41,4

9,2

65,4 68,3

0,0

14,7

30,9

51,0 53,3 57,8

Env.

Crop Growth

19,9 25,2

44,4 48,1

16,4

P11 0,0

11,9 16,7

18,7 32,2 38,7

161,1 170,4

83,4

0,0 10,3

24,5

38,5

148,2 156,3

174,1

P8

P6 0,0 2,8 12,6

Frd12.1IM

27,7 31,5 32,4 37,4 45,8 49,1 53,7 55,6 57,7

params

Challenging: • high plants (3 m high) • row space small(< 60 cm) • intertwined plants

- 530 markers (AFLP, RFLP, SSR, KG,...) assigned to 12 chr. (1500 cM) - 100 markers / 20 small LG (+ 300cM)

7,1 11,7 16,6 21,3

params Generic

Phenotyping large pepper plants

Genetic map

P1

WUR (coordinator), INRA (FR), VIB (BE), James Hutton Institute (UK), Budapest Univ (HU), Exp. station Cajamar (ES)

Genetic

83,1 85,4

49,6 52,8

50,2 56,0 61,0

56,1 60,7 67,9 71,1

67,1 72,4 81,8

74,0

74,8 78,5 81,4 90,3 97,6 104,8

117,5 124,4

4 RGB camera’s 4 infra-red camera’s 4 range camera’s Use wide angle lenses

83,0 205,4

205,2

5


10/19/2011

Problem: varying lighting conditions

System setup

IR camera

RGB camera Range camera (TOF)

Flash light

Record images every 5 cm

Rig in greenhouse

Mirror

Time-Of-Flight Range camera   

RGB image second row Infrared

second row

A Time-of-Flight (ToF) camera can produce a depth image. The camera illuminates the scene by infrared light. Distance in cm is calculated from the time the light has used for travelling to the object and back. Low resolution

QR barcode

Approach 

Reconstruct 3D canopy (depth information)

Then extract features        

number of leaves size of leaves leaf area angle of inclination of leaves number and size of fruits stem thickness internode length ....

3D canopy reconstruction 

 

Combine stereo color images and ToF range image by mapping them to a single 3D reference coordinate system ToF image: (mixed) pixel represents a patch in 3D Integrate ToF image with stereo RGB images using graph cuts (GC). Yu Song, C. Glasbey, G.W.A.M. van der Heijden G. Polder and A. Dieleman. Combining stereo and Time-of-Flight images with application to automatic plant phenotyping. SCIA. 2011.

6


10/19/2011

Depth estimation using stereovision

Coarse ToF image

Disparity

Depth

Pinhole camera model:

z=sf/d d1

Low resolution ToF image

Pre-defined parameters s and f

Combine with highresolution RGB images Different viewing position and lens

Automated estimation

d2

Dense correspondence (depth for every pixel in image)

Results RGB

Measuring leaf area SIFTflow

SHAPE

GC

ToF

ToF+GC ď Ž

Now we have a 3D reconstructed scene, we can automatically extract a leaf from the scene and compute its surface area.

Depth estimation using 3 state-of-the-art stereo algorithms (SIFTflow,Shape,GC), using only ToF and using our method, combining ToF and GC. Yu Song, C. Glasbey, G.W.A.M. van der Heijden G. Polder and A. Dieleman. Proceedings of SCIA. 2011.

Smoothing of leaf surface

Leaf area results Leaf area (cm2)

Plant 1

Plant 2

Plant 3

Manually measured

67.25

97.27

36.60

No smoothing

109.66

149.38

62.65

Smoothing

69.70

102.40

33.74

7


10/19/2011

High Throughput and Deep Phenotyping

Data explosion

VIS/NIR/UV/ X-ray/NMR/... metabolomics proteomics

phenotyping depth

 standard measurements

when (development in time)

where (cell/tissue/organ) conditions (different enviroments)

We generate far too much data to handle manually Simple summary statistics as means and standard deviations do not suffice Advanced analysis tools are required

transcriptomics

genotyping depth

full genome sequencing

An example: QTLxE analysis Barley

Phenotypic scores

P1

DH1

Effects of QTL on chromosome Population of e.g. Doubled Haploids, Genetic information

P2

DH2

DH…

DH n

P1 Allele Superior P2 Allele Superior No significant effect

Environments

Environmental information: climate, soil

3.4

4.4

3.4

….

1.0

3.4

3.5

2.4

….

2.0

3.1

3.5

2.6

….

4.0

2.8

3.0

2.4

….

3.0

2004

QTL-effects related to temperature

2005 Boer, M.P.; Wright, D.; Feng, L.; Podlich, D.W.; Luo, L.; Cooper, M.; Eeuwijk, F.A. van. Genetics 177 (3). - p. 1801 - 1813. 2007.

Thank you for your attention Contact: gerie.vanderheijden@wur.nl © Wageningen UR

8


New Genotype To Phenotype Models At The Intersection Of Genetics, Physiology And Statistics; Smart T