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

10/19/2011

1

Plant Phenotyping at Wageningen UR

Gerie van der Heijden ([email protected])

Phenodays, WageningenOctober 13, 2011

From Genotype to Phenotype

Improved crops

Genotype……………………………...Phenotype

Crops

Environmental variation

Seq

ue

nc

ing

+

Ge

no

typ

ing

Ph

1

42

3

5

Phenotyping at Wageningen UR

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

Image analysis and automation

Vision system for sorting of orchids

Contact: Rick vd Zedde

Automatic harvesting of roses

Contact: Rick vd Zedde

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

10/19/2011

2

Plantalyser

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

Monitoring Arabidopsis

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

Hyperspectral imaging

Acquisition of hyperspectral images Hyperspectral imaging of tomatoes

Tomatoes in different ripeness stadia

10/19/2011

3

Measurement of Lycopene content

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.

HPLC measurement

Detecting Fusarium in wheat seeds

A. color reflection

B. NIR transmission at 1100 nm

C. Predicted Fusariumconcentration

Use of transmission NIR (900-1800 nm)

15.5 16 16.5 17 17.5 18 18.5 19 19.5 20 20.513

14

15

16

17

18

19

20

21

--102

--104

--108

--110

--202

--203

--205--206

--208

--210

--211

--212

--301

--302--304

--305

--307

--308

--309

--310

--311

--312

--401

--403--407

--408

--409

--411

--412

--501

--503

--505

--507

--508--509

--510

--511

--512

--601

--602--603

--605

--606--610

--611--702

--703--706

--707

--708--710--711--802

--805--807--808

--809--810

--812

Pixel error:Object error:RMSEP: 0.82

RMSEP/: 0.04

Q2: 0.79

RMSEP: 0.79

RMSEP/: 0.04

Q2: 0.80

Detecting Fusarium in wheat seeds

Measured Ct

Pred

icte

d Ct

PLS with cross validation on seed kernel basis

G. Polder, G.W.A.M. van der Heijden, C. Waalwijk, and I.T. Young. Seed Science and

Technology. 33(3):655–668, 2005.

Hyperspectral imaging of grassland

velocity 0.3-0.5 m/s

1

2

3

4

5

1 2 3 4 5

Voorspeld N (%)

5

10

15

20

25

30

5 10 15 20 25 30

DS gehalte (%)

0

1

2

3

4

0 1 2 3 4DS opbrengst (ton ha- 1)

Predict biomass, DMC and N-content

DM yield (ton/ha) DM content (%) N content (%)

Mea

sure

d

Predicted

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.

Combining remote and close sensingPhysical and

chemical measurement

Close sensing

Remote sensing

cal cal

10/19/2011

4

Biomass and nitrogen prediction in the field

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.

CF Transient Imager

Contact: Henk Jalink

F0

Fmax

Fv/Fm = (Fmax– F0)/Fmax

Imaging the induction curve of photosynthesis

Quantification of stress in leaves

G. Polder, G.W.A.M. van der Heijden, H. Jalink and J. Snel. Computers and electronics in Agriculture, 55:1-15, 2007.

Time sequence cabbage plants

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

10/19/2011

5

Control 6D Salt 13D Salt

Sensitive

TolerantTraits:

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

Find responsible genomicregion: QTL

Contact: Henk Jalink and Gerard vd Linden

Salinity stress in potato High-throughput 3D seedling sorter

Aim: Seedling assessment based on human

expert knowledge modelling

Approach: Highly accurate 3D reconstruction using

volumetric intersection of 10 camerassimultaneously

Result: a high-speed sorting device in cooperation

with Flier Systems BV.

High-throughput 3D-based seedling sorting

Capacity: 20.000 seedlings/ hour.

Processing time: 45ms per seedling

Contact: Rick van de Zedde

Plant Phenotyping in EU project SPICY

Combine phenotypic data with genotypic data and crop growth models for pepper

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

QTLGenetic

params

Generic

params

Env.

params

Crop

Growth

Model

Pheno-

type

Recombinant inbred linesof pepper

Yolo Wonder CM334

genotyping 297 RIL

F5YC

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

Genetic map

0,0

17,8

18,5

22,2

36,3

40,7

53,6

62,4

67,1

70,2

83,4

91,1

93,6

95,9

101,5

104,6

110,7

P2

0,0

30,8

38,6

42,4

44,5

49,3

56,4

63,4

70,1

76,6

80,9

85,3

90,6

93,8

102,0

111,8

148,2

156,3

161,1

170,4

172,9

P4

0,0

9,0

16,4

20,6

24,4

28,3

31,2

34,1

36,1

38,5

45,1

54,4

83,0

P7

0,0

10,3

13,2

18,7

35,2

40,0

45,3

47,5

52,9

55,7

86,7

P8

0,0

9,2

20,3

28,9

43,9

50,2

56,0

61,0

67,1

72,4

81,8

0,0

11,9

16,7

25,2

34,6

P100,0

7,8

14,7

19,9

25,6

0,0

19,6

28,8

31,7

33,0

35,3

38,0

41,5

45,1

46,8

49,6

52,8

56,1

60,7

67,9

71,1

74,0

74,8

78,5

81,4

90,3

97,6

104,8

117,5

124,4

P110,0

6,5

11,4

13,2

23,1

30,9

40,2

44,0

46,9

54,9

62,0

83,1

85,4

Frd12.1IM

P12P10,0

7,3

15,0

19,1

28,2

33,3

35,3

37,2

41,8

51,655,8

60,1

62,2

66,9

69,9

72,3

74,3

89,5

97,6

101,9

104,0106,1

108,5

112,3

114,2

116,7119,4

123,7

128,7

134,6

137,7

142,6

145,2

148,7

180,8

184,9

188,6

192,4

197,8

205,4

P50,0

7,3

15,0

19,1

28,2

33,3

35,3

37,2

41,8

51,655,8

60,1

62,2

66,9

69,9

72,3

74,3

89,5

97,6

101,9

104,0106,1

108,5

112,3

114,2

116,7119,4

123,7

128,7

134,6

137,7

142,6

145,2

148,7

180,8

184,9

188,6

192,4

197,8

205,4

P50,0

7,3

15,0

19,1

28,2

33,3

35,3

37,2

41,8

51,655,8

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

116,7119,4

123,7

128,7

134,6

137,7

142,6

145,2

148,7

180,8

184,9

188,6

192,4

197,8

205,4

0,0

2,8

12,6

24,5

32,2

38,7

44,4

48,1

51,0

53,3

57,8

60,4

62,5

65,4

68,3

70,7

74,8

P60,0

2,8

12,6

24,5

32,2

38,7

44,4

48,1

51,0

53,3

57,8

60,4

62,5

65,4

68,3

70,7

74,8

P60,0

8,8

13,7

41,4

55,6

76,7

90,4

93,4

100,1

107,0

108,3109,5

110,6

111,9

114,6115,7

119,2

121,2

124,3

128,3

132,1133,7

136,2

139,6

144,0

148,7

154,0

157,1

168,5

180,5

185,3

189,5

194,0

205,2

0,0

8,8

13,7

41,4

55,6

76,7

90,4

93,4

100,1

107,0

108,3

109,5

110,6

111,9

114,6115,7

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

205,2

P9P30,0

7,1

11,7

16,6

21,3

27,7

31,5

32,437,4

45,8

49,1

53,7

55,657,7

65,7

95,7

130,5

137,9

146,4

153,0

160,7

174,1

SPICY: Plant material and genetic map Phenotyping large pepper plants

Build high rig with: 4 RGB camera’s 4 infra-red camera’s 4 range camera’s Use wide angle lenses

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

10/19/2011

6

Problem: varying lighting conditions System setup

Rig in greenhouseIR camera RGB camera Range camera (TOF)

Flash light Mirror

Record images every 5 cm

RGB image second row Infrared second rowQR barcode

Time-Of-Flight Range camera

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

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.

10/19/2011

7

Depth estimation using stereovision

Disparity Depth

Pinhole camera model:

z = s f / d

Pre-defined parameters s and f

Automated estimation

Dense correspondence (depth for every pixel in image)

d1

d2

Coarse ToF image

Low resolution ToFimage

Combine with high-resolution RGB images

Different viewingposition and lens

Results

Depth estimation using 3 state-of-the-art stereo algorithms (SIFTflow,Shape,GC), using only ToF and using our method, combining ToF and GC.

RGB SIFTflow SHAPE GC ToF ToF+GC

Yu Song, C. Glasbey, G.W.A.M. van der Heijden G. Polder and A. Dieleman. Proceedings of SCIA. 2011.

Measuring leaf area

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

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

10/19/2011

8

High Throughput and Deep Phenotyping

genotyping depth full genome sequencing

when (development in time)

where (cell/tissue/organ)

conditions (different enviroments)

transcriptomics

standardmeasurements

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

proteomics

metabolomics

phen

otyp

ing

dept

hData explosion

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

An example: QTLxE analysis

P1 P2

DH n

3.4 4.4 3.4 …. 1.03.4 3.5 2.4 …. 2.03.1 3.5 2.6 …. 4.02.8 3.0 2.4 …. 3.0E

nvir

onm

ents

2004

2005

Population of e.g. Doubled Haploids, Genetic information

DH1 DH2 DH…

Barley

Environmental information: climate, soil

Phenotypic scores

Effects of QTL on chromosome

P1 Allele Superior

P2 Allele SuperiorNo significant effect

QTL-effects related to

temperature

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

© Wageningen UR

Contact: [email protected]