Download Phenotyping of E18.5 Mouse embryos using MRC

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Phenotyping of E18.5 Mouse embryos using MRC Harwell’s automated morphological phenotyping pipeline
Authors: T. Barton, N. Horner, J. Brown, H. Westerberg.
Introduction
The Medical Research Council Harwell (MRC Harwell) is part of the International Mouse Phenotyping Consortium (IMPC) whose goal is to phenotype 20,000 mice genes by the end of 2021. It is estimated
30% of mice genes when knocked out are embryonic or perinatal lethal (non-viable), of this half are lethal between E14.5 and E18.5. At Harwell this class of lines are scanned using Micro CT and
processed through an automated phenotyping pipeline to find morphological abnormalities. (~ 75 lines ). This project involved using MRC Harwell’s Automated Phenotyping Pipeline (HAMPP) on E18.5
mice embryos (18.5 days old embryos), and developing the registration pipeline.
Embryo pre-processing
What is an automated morphological phenotyping pipeline?
Automated morphological phenotyping uses computer image processing for comparing 3D Micro CT scanned mouse embryos are pre-processed in NRecon1 to remove noise, correct alignment and
images (volumes) of animals to detect significant morphological (anatomical) differences, reduce ring artefacts and reconstructs to a 3D image, then input into the Harwell Automated Recon Processer
the pipeline covers the processes involved in this. Diagram. 1 shows the basic layout out of (HARP) which crops, scales, and stacks the images in single file volume.
the pipeline.
Preprocessing
Non-viable or sub viable
embryos
Micro CT Scan
Registration
pipeline
NRecon
HARP
Embryo Volumes
Analysis
Image. 2– Sagittal slice view of
E18.5 population average 3D
Slicer3
Moving Volumes
Image. 1 – Example of embryo in NRecon (above)
Diagram. 2 – Flow diagram of preDiagram. 1 – Flow diagram of
processing
HAMPP (Above)
Registration pipeline and phenotype detection
First, a large sample of wild type embryos are input into the registration pipeline to
create a population average, see diagram 3. Then a new sample of wild type
embryos are registered with the population average as the fixed volume for each
registration step, and an analysis of the transform vectors of the non-rigid
registration step.
a) Fixed
a) Moving
Finally, a set of mutant embryos with a specific gene knocked out are registered
with - the population average as the fixed volume for each registration step. An
analysis of the transform vectors of the non-rigid registrations step, and registered
wild type and mutant is carried out.
A t-test, with filtering for false detection rate, is carried out on the comparison so
only differences between the wild type and mutant embryo volumes which are
statistically significance are output on to a heat map where hits represent
a) Affine
a) Rigid
a) Non-Rigid
morphological differences in the mutant embryos.
I used seventeen wild type embryos to create the population average. Seven wild Image. 4 - Effects of different registration transforms on a image4
types, and three mutant embryos of same the line were used for wild type to Rigid – translation and rotation transforms
mutant comparison. During the course of the project the population average was Affine – Rigid with scaling and shearing transforms
corrected for distortion due to the non-rigid registration step, and improper Non-rigid – deformation of grid on the moving image to fit the
alignment. I was able to process a number of mutants, the NTRK1 mutant appears same grid on the fixed image
below as an example.
Registration: transforming a moving image on to a fixed image,
Image. 3 3D rendering of E18.5
population average in 3D Slicer3
Fixed Volume
Rigid
Registration
Rigid Registered
(moving) Volumes
Rigid Average
(fixed) Volume
Affine
Registration
Affine Registered
(moving) Volumes
Affine Average
(fixed) Volume
7 Wild Types
Non-Rigid
Registration
Non-Rigid
Registered (fixed)
Volume
Non-Rigid Average
Volume (Population
Average)
Registration
Pipeline
3 MUT/HOM
Registration
Pipeline
Diagram. 4 – Showing the
pipeline for mutant vs.
wildtype embryo comparison.
carried out using the program elastix2.
Intensity heat map
Comparison &
Statistics
Deformation heat map
Determinant of the Jacobian heat map
Population Average (fixed
volume)
Diagram. 3 – Flow diagram of registration
pipeline, for population average
NTRK1 Gene analysis
Image 10 & 11 – Showing 3 mutants (top) embryos vs. 3 wild type
(bottom) embryos comparison.
Image. 6 & 7– Shows a determinant of
the Jacobian heat map overlaid onto a
registered NTRK1 mutant. The blue
highlight region are areas of expansion,
this region is the trigeminal ganglia. The
left image is axial slice view, and the right
is a coronal view, viewed in Harwell
Volume Viewer.
Image. 8 & 9 – Showing a manual measurement of trigeminal ganglia on a
NTRK1 HOM embryo (right) and wild type embryo (left). Visually there is a
difference between the NTRK1 mutant and wild type, however for more
quantitative results manual measurements were taken for 3 mutant and 3
wild type, the results can be seen in table 1.
Trigeminal ganglia measurement from end to end
Litter
Left Length/mm
Right Length/mm
NTRK1
Average
Wild Type
Average
16.2f 13.3j 13.3h
22.3b 17.2h 15.3j
1.36
1.37
1.15
1.29
1.69
1.79
1.55
1.67
1.42
1.35
1.22
1.33
1.53
1.78
1.62
Table. 1 – Showing trigeminal ganglia longest point to point length
measurements
1.64
Summary
The HAMPP was able to create an E18.5 mouse embryo population average and generate heat maps for mutant to wild type comparison. Analysis was carried out a number of genes to
show statistically significant morphological changes in the mutant embryos which could be used by developmental biologists for researching phenotypes of knockout genes. In order to
develop the registration pipeline Python code was written to test different registration parameters for optimisation. This project developed my knowledge of Biophysics, Bioinformatics,
image processing, image visualisation, Python programming, bash programming and Genetics.
Acknowledgements
I would like to thank MRC Harwell, SEPnet and Queen Mary, University of London for funding this internship. Also, would like to thank every at MRC Harwell who supported me
throughout this project, in particular the members of SIG research; Henrik Westerberg, Neil Horner, and James Brown, and the Bioinformatics team.
References
1) NRecon - http://bruker-microct.com/products/downloads.htm
2) Elastix - http://elastix.isi.uu.nl/
3) 3D Slicer - http://www.slicer.org/
4) Elastix the manual S. Klein, M. Staring, February 12, 2014.