My Masters Work

1
My Masters Work

• Richa Tiwari

2
Outline of the talk

• Analysis of Phylogeny Tree Evaluation Approaches
  (Project done in CS641).
• Proteomics and 2-D Gel Electrophoresis (Study
  done for CS)
• Coexpression analysis of dimerization between
  bZIP proteins in groups C, S1 and S2 in
  Arabidopsis Thaliana, under the conditions of
  differential light and CO2 levels (Project done
  for BST676).

3
Analysis of Phylogeny Tree Evaluation Approaches
4
Phylogenetic Analysis

• Alignment of the sequences
• Determining the presence of relationship between
  sequences
• Decision of most appropriate tree building
  algorithm
• Scrutinize the tree to determine level of
  confidence

5

• Algorithmic Method
• Defines an algorithm that leads to the
  determination of a tree.
• Criteria Based Method
• Defines a criterion for comparing different
  phylogenies and therefore phylogenies can be
  ranked, and comparison possible.

6
(No Transcript)
7
Maximum Parsimony Method

• Most parsimonious tree will explain the observed
  character distribution with a tree that have the
  minimum tree length.
• Tree selection criterion - Minimum tree length
• (Fewest character state transformation)

8
Maximum Likelihood (ML)

• ML evaluates the probability that the chosen
  evolutionary model will have generated the
  observed sequences.
• Evolutionary Model Accounts for the changes in
  sequences.
• Phylogenies are then inferred by finding those
  trees that yield the highest likelihood.

9
Distance Based Method

• Distance-based methods attempts to find the
  distance that is the total changes between the
  two taxons from the point where they last shared
  an ancestor.
• It is a cluster based method.

10
Software used.

• PHYLIP
• To compare the three phylogeny methods.
  Programs used from the package are
• Maximum Parsimony DNAPARS
• Maximum Likelihood DNAML
• Distance-based DNADIST and Neighbor
• Tree constructed using DRAWGRAM
• Consensus tree constructed using CONSENSUS

11
Using Sample data
Maximum parsimony
Maximum likelihood
Distance Based

DNAPARS DNAML Neighbor
12
Consensus tree for given example
------Human
--1.0- ------Orang
------ ------Rhesus
--1.0-
------Gorilla --------------------Chimp
------Human
--1.0- --1.0-
------Chimp ------
-------------Orang
--------------------Rhesus
---------------------------Gorilla
Parsimony Method
Maximum Likelihood
-------------Orang
--1.0-
------Chimp ------ --1.0-
------Human
--------------------Rhesus
---------------------------Gorilla
Distance Based/Neighbor joining
13
Observation

• Reliability of branch length estimates
• NJ and MLgt MP
• Computational speed (ngt500)
• NJ/DNADIST 0.005 seconds
• DNAPARS 0.5 seconds
• DNAML 230.0 seconds

14
Conclusion

• Our experiments and the results obtained indicate
  that the Distance Based method is better than the
  other two methods in terms of Fastness,
  Simplicity and good performance for high number
  of taxa.
• Also we can say that if you have a fast computer
  and large dataset Maximum likelihood method is
  better than Maximum parsimony.

15
Proteomics and 2-D gel Electrophoresis
16
Introduction

• The entire set of proteins expressed by the
  genome in a cell, organ or organism is referred
  to as the proteome.
• Proteomics Methods that discover and quantify
  proteins and their biochemical changes.

17
Application of Proteomics

• Protein Mining
• Network Mapping
• Mapping Protein Modifications

18
Proteomics Analysis
Reference www.mbi.osu.edu/sciprograms/prfmaterial
s/vandre.ppt
19
2-D Gel Electrophoresis

• The horizontal position tells us about the charge
  of a protein, whereas the intensity of the gel
  spot tells us about the amount of that protein in
  the system.
• Steps-
• 1. Prepare protein sample in solution
• 2. Separate proteins (in each dimension)
• I. Based on pH
• Using isoelectric focusing (IEF)
• Using immobilized pH gradient (IPG) strips
• II. Based on molecular weight (size)
• Using gel electrophoresis
• 3. Stain proteins to enable visualization.

20
Introduction to the project

• This project focuses on 2D gel electrophoretic
  separation of proteins.
• We analyzed few random spots from the 2D gels of
  rat mammary tissue.
• Statistical methods to find the variance in pI of
  the same protein in different gels.
• Analyzed the reasons for these differences.
• Inferred the relationship between the
  experimental values and the predicted values.

21
Images of the gels used in the project.
22
One of the gels with Protein Spots
23

• The Gels we used were from an already done
  experiment. 28 Random protein spots were selected
  based on the their intensity from each of the
  three gels.
• Mass Spectrometry
• Differentially expressed proteins identified by
  image analysis were excised from 2D gels and
  trypsin digested. The resulting peptide fragments
  were analyzed on a MALDI mass spectrometer (MS).
  The MALDI spectra displays a peptide
  fingerprint of the protein using corresponding
  peptide masses.

24
MALDI TOF MS
25

• Proteins were identified by entering the masses
  (ions from MALDI spectrum) of the peptides into
  a peptide mapping database. Some examples of such
  protein search engine are-
• Mascot - very popular and also used in this
  project
• Sequest
• Aldente
• ProteinLynx
• Phenyx

26
Image of a search data base
27
Results

• We tabulated the result obtained from the
  database internet search and the one we obtained
  from the experiment.
• We observed that the pI values as well as the
  molecular weight were not same in all gels for
  same protein.
• The pI values of the three gels were quite
  similar but they were different from the
  predicted pI values.

28

• In a 2D gel the position of protein spot can
  change due to various reasons and because of
  which the molecular weight and pI values may also
  differ.

29
Graphical representation of pI values of three
gels
30
Graph showing the variance among the predicted pI
and observed pI
31
Observations

• We saw that the difference between the pI values
  of the three gels that is the experimental values
  are not very different from each other.
• So we can interpret that the difference due to
  non biological reason is very less in the
  experiment.
• There were few protein spots for which internet
  search revealed the same result as same protein
  name. But our experiment gave different results
  which can be because of different group (like
  phosphate or sulphate) getting attached to it.
  There can be other reasons for it too.

32

• Average deviations between the three observed
  proteins and the predicted pI values were
  calculated as
• (pI (gel 12_5)- pred. pI) (pI (gel 12_5)-
  pred. pI) (pI (gel 12_5)- pred. pI) / 3
• This gave the results shown in the next slide.
  We obtained positive as well as negative values
  for the deviations.

33
Average deviations between the three gels and the
predicted pI
34

• We can interpret that the proteins were modified
  more by negatively charged group such that there
  pI values decreased.
• The addition of one phosphate groups to serine,
  threonine, and tyrosine residues typically
  decreases their isoelectric points by 0.1 pH
  unit.

35
Regression results

• A statistical analysis test was performed to
  determine which of the three gels were closest to
  the predicted pI values. That is in which of the
  three gels had the proteins being least modified.
• The test was Clibration test. We prepared a
  regression model for each gel. The inverse
  regression equation used was
• Predicted pI Observed pI from Gel Intercept
  slope

36
Predicted pI values from the Calibration test and
internet database
37

• The result we obtained showed us that all the
  three gels predicted almost same pI values and
  they were quite away from the original predicted
  pI values.
• All these similarities between the three gels
  show us that the difference between the pI values
  of proteins between the predicted and the
  experimented values is not very much because of
  non biological factors, but because of chemical
  modifications in the proteins.

38
Coexpression analysis of dimerization between
bZIP proteins in groups C, S1 and S2 in
Arabidopsis Thaliana, under the conditions of
differential light and CO2 levels.
39
IntroductionTranscription factor

• Transcription factor are proteins involved in the
  regulation of gene expression, that bind to
  promoter region upstream of genes.
• They are composed of two essential functional
  regions
• DNA binding domain It binds to DNA.
• Activator Domain It interacts with other
  regulatory proteins there by affecting the
  efficiency of DNA binding.

40
bZIP proteins

• bZIP proteins are a class of transcription factor
  which has leucine zipper motif consisting of a
  periodic repetition of a leucine residue at every
  seventh position forming an alpha-helical
  confirmation.
• The segment that comprises the basic region and
  the periodic array of leucine residues is
  referred to as basic-region leucine zipper or
  bZIP motif.

41
(No Transcript)
42
Some facts

• There are 792 bZIP proteins recorded in
  nonredundant database.
• The no of bZIP proteins in the cell of selected
  organisms are as follows
• yeast 16
• fruitfly 110
• plant (Arabidopsis thaliana) 75
• Human - 114

43
Arabidopsis

• The Arabidopsis genome sequence contains 75
  distinct members of the bZIP family, of which 50
  of them are not well studied.
• Using common domains the bZIP family can be
  subdivided into 10 groups Groups A - S.

44
(No Transcript)
45
(No Transcript)
46
C S protein interaction

• Elhert et al measured interactions between C and
  S proteins.
• C and S1 heterodimerized
• Two S2 proteins dimerized.

47
Effect of Light CO2 on C S proteins

• Carbohydrate signaling
• Increase of carbohydrate partitioning in
  elevated CO2, and a decrease in low light.
• Seed development
• Photosensory system detects the quality,
  quantity, direction and duration of light.
  Controls developmental pattern.
• Stress
• Light dependent generation of active oxygen
  species is a type of stress called photo
  oxidative stress.

48
Experiment Selection Criteria

• a) Chose C and S bZIP proteins
• Coexpression Engine http//www.ssg.uab.edu/coexpr
  ession
• b) Selected tissue and array type
• c) Chose specific experiment

49
a) Chose C and S bZIP proteins
50
b) Selected tissue and array type
51
c) Chose specific experimentNASC Experiments
52
Justification

• Biologically feasible comparisons due to similar
• Tissue types
• Experiment conditions
• Statistical
• Measurement protocol

53
The tool used

• Co-expression Analysis Tool, version 2.0
  developed at the Section on Statistical Genetics,
  UAB http//obiwan.ssg.uab.edu8080/coexpression/se
  rvlets/CoexpReleasesResponseManager
• mainly built to analyze the co-expression in
  Arabidopsis plant.
• NASC Experiments to study affymetrix gene chip
  profiling of light and CO2 effect in leaf
  development in Arabidopsis used.

54

• Uses the database built from Nottingham
  Arabidopsis Stock Center (NASC) AffyWatch
  Service.
• Version 2 used in this project contains total of
  566 microarray chips out of which 486 ATH1 micro
  array chips were used.

55
NASC Experiments used

• 4 experiments conducted to examine the effect of
  developing leaf insertions under varying
  conditions of light and CO2.
• The sampling was done at time interval of 0th,
  2nd, 4th, 12th, 24th, 48th and 96th hour using a
  batch of 24 plants.
• Four replicates were produced for each of the
  seven time points per experiment.

56
Working of the tool

• Linear regression analysis is done on the probe
  sets.
• Result of regression gives three important
  values- slope parameter (indicating the direction
  of co-expression), p-value (stating the
  confidence in the correlation) and R squared
  values (strength of correlation).

57
Procedure

• 4 genes of C group, 5 genes of S1 group and 3
  genes of S2 group were studied in the project.
• We submit the AGI IDs, the tissue type (here
  leaf) and the experiment number (in our case 156,
  157 158 and 159) in the tool.
• Our genes of interest are regressed on all the
  22,810 ATH1 probe sets and a p-value, R squared
  value and slope parameter is obtained.

58

• Those genes were subsequently sorted according to
  the R squared value and p-value and ranked such
  that
• Higher the R squared value, higher is the
  rank.
• An arbitrary cut-off 15 of the top ranked genes
  were identified as highly co-expressed.

59
Hypothesis

• Genes coding for dimerizing proteins should be
  coexpressed at the same time.
• If genes in group C and S1 lead to
  heterodimerization then they should be
  coexpressed at the same time.

60
Table 2 Mapping information between AtbZIP AGI
ATH Probeset AtbZIP Group Ids
61
Table 3 Regression estimates between Group C
AtbZIIP63 (245925_at) and Probes in Group S1, C
and S2.
62
Table 4 Regression estimates between Group C
AtbZIIP25 (251848_at) and Probes in Group S1, C
and S2.
63
Regression estimates between Group C AtbZIIP9
(246962_s_at) and Probes in Group S1, C and S2.
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
(No Transcript)
68
(No Transcript)
69
(No Transcript)
70
(No Transcript)
71
Results

• bZIP1(Group S1) coexpresses well with bZIP63
  (S1) under conditions of Ambient Co2 and low
  light but the same coexpression interaction is
  weak under conditions of Elevated Co2 and Ambient
  Light.
• Also, very minimal interaction was found between
  genes of Group C (bZIP25, bZIP10, bZIP9, and
  bZIP63) and bZIP9 (Group C

72
Conclusion

• This bZIP study was a good litmus test for the
  SSG Coexpression Tool.
• Results presented in this study provide evidence
  that a good if not significant number of AtbZIP
  proteins interacting as heterodimers are
  co-regulating under varying conditions of stress.
  
• This study shows evidence that coexpression
  patterns in genes can be studied by pooling
  publicly available microarray data and that the
  use of simple linear regression procedure is
  feasible.

73
Discussion

• Varying trends in the coexpression proposes some
  theories
• Different genes are expressed in diff tissues. Is
  study on leaf good enough to support our
  hypothesis?
• Time-course data is valuable and should be
  accounted for in the analysis. However, this kind
  of analysis requires more observation recorded at
  different timepoints.
• Linear regression is good but will a robust
  time-series based approach be appropriate in our
  study?