Transcript FTIR data analysis tutorial

FTIR data analysis tutorial
Bryan Penning
*Supported
bythe
theNSF
NSF
Supported by
Plant
Genome
Research
Plant Genome Researchand
REU
and
REUPrograms
Programs
Overview
• We are establishing infrared “spectrotypes” in cell wall
mutants. Spectrotypes are spectroscopic phenotypes, i.e.
the spectral differences between mutant and wild type
populations.
• This tutorial will:
– Give some background information on how we prepare and
collect data on mutants (See Techniques VII at:
http://cellwall.genomics.purdue.edu/techniques/7.html for more
details)
– Show how we analyze the data collected by Principal
Components Analysis (PCA) and digital subtraction
– Give some example spectral peaks that indicate differences in
cell wall composition between mutants and wildtype
Supported by the NSF
Plant Genome Research
and REU Programs
Plant preparation
• All plants are grown and cell walls isolated under the same
conditions and processed simultaneously for internal
consistency of mutant and wild type comparisons (see
Techniques VII for greater detail:
http://cellwall.genomics.purdue.edu/techniques/7.html)
– Wild type plants are always prepared with mutant plants
– Plants are all grown on one-half strength MS salts with 1%
sucrose and 0.8% agar in the light for 12 days
– Plants are all transferred to the dark for 2 days to lower
starch content (a contaminating feature in IR spectrum)
– Plant tissue is crushed in liquid nitrogen, and non cell wall
components, such as proteins, are extracted using an SDS-Tris
buffer
– Cell walls are isolated by homogenization in a Geno-Grinder
(SPEX Certi-Prep), collected on a nylon mesh and washed with
water and ethanol
– Cell walls are resuspended in distilled water and spotted on a
gold-plated slide (EZ-Spot, Spectra-Tech) for spectral
acquisition in an FTIR spectrometer (Thermo-Electron,
Madison, WI )
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR spectral acquisition
• Spectra are acquired in a range of 4000 to 650 cm–1; 1
spectrum consists of 128 co-added scans with an eightwavenumber resolution
• The following spectral characteristics are used to ensure that
readings can be compared:
– 0.4 to 0.8 max peak reading @ 1050-1000 cm-1
– Peak @ 1050-1000 cm-1 greater than peak @ 1600 cm-1
– The values @ 1800 cm-1 and 800 cm-1 are fairly equal with a value
below 0.3 absorbance (to ensure a good baseline correction)
– Noisy spectra are discarded
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
A
B
• Fig. A: We collect many spectra (about 40) with a computer driven
stage and Omnic software (Thermo Electron), saving them as
grouped data (*.spa)
• Fig. B: We convert group data to individual *.jdx files in Omnic
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
•We use Win-Das (Kemsley, 1998)* software to analyze our data
instead of Omnic because it is capable of area averaging the spectra,
an essential feature of spectral analysis of plant cell walls because the
samples vary in thickness
•With the DOS
command, RENAME, we
convert *.jdx files into
*.dx files that Win-Das
software can recognize
* Kemsley. 1998.
Discriminant Analysis of
Spectroscopic Data.
Chichester, UK: John Wiley
and Sons
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
• In Win-Das we first construct a
matrix
• We add all of our spectra
(observations)
• We view the spectra
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
• In Win-Das we:
– Truncate the spectra from 1801.2
to 798.4 cm-1 (useful wavenumber
range for cell wall molecules)
– Baseline correct the spectra
– Normalize the spectra and save as a
.txt (spectra) or .wdd (analysis) file
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
• Saving data for the web
– After normalizing we save the file
as CF-Text (column-wise)
– This generates a *.txt file we copy
and paste into Excel
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
•
•
Supported by the NSF
Plant Genome Research
and REU Programs
You can download these spectra
from our website (Families Tables)
You can average the spectra using
the average command in Excel
FTIR data analysis
• Digital subtraction
– To perform a
digital subtraction,
average the mutant
and wild-type
spectra (previous
slide) and copy over
the spectra values
(left column of
spectra files)
– Subtract mutant
from wild type and
plot versus
wavenumber (cm-1)
– Look for peaks
(differences in cell
wall components)
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
• Absorbances of specific peaks in the IR spectrum can be
correlated with particular cell wall molecules (Kačuráková et al,
2000)*:
–
–
–
–
Cellulose: 1162, 1120, 1059, 1033, 930, and 898 cm-1
Pectin: 1144, 1100, 1047, 1017, 953, 896 cm -1
Rhamnogalacturonan: 1150, 1122, 1070, 1043, 989, 951, 916, 902 cm -1
Xyloglucan: 1153, 1118, 1078, 1041, 945, 897 cm -1
• However, these peak assignments are based on isolated
polysaccharides and peaks may shift depending on molecular
interactions and environment within the cell wall
• The more peaks that can be assigned to a particular polymer, the
more likely that component differs between mutant and wild type
cell walls
* Kačuráková, Capek, Sasinková, Wellner, and Ebringerová. 2000. FT-IR study of plant
cell wall model compounds: pectic polysaccharides and hemicelluloses. Carbohydrate
Polymers 43:195- 203
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
• To develop Discriminant Analysis for
our classifications (PCAs) in Win-Das
– We create two groups (one wild type
and one mutant)
– We compress the data using the
covariance method
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
• PCA analysis:
– Cluster plot of separation by PCAs
– Loading plots (difference in groups)
• Discriminate
– We use Squared Mahalanobis
distance to see number of correct
classifications
Supported by the NSF
Plant Genome Research
and REU Programs
FTIR data analysis
• PCA from Variance scores are shown
in the gene family table…
– Number of Principal Components (PCs)
– Percent classified (75/80 = 93%)
Supported by the NSF
Plant Genome Research
and REU Programs