Download UlrikPhD2005 - Center for Biological Sequence Analysis

CBS Ph.D. course, 14. april 2005
Using
Microarrays
to Study
Cell Cycle
Regulation
Ulrik de Lichtenberg
Center for Biological Sequence Analysis
The Technical University of Denmark
Outline
• Introduction to cell division and the cell cycle
• Cell culture synchronization and arrest methods
• Microarrays and periodically expressed genes
• Our own research
The Origin of Life
How does one cell become many?
The cell cycle
The cell cycle – a hot scientific topic
9000
8000
7000
The 2001 Nobel Prize in
Physiology or Medicine
Annual number of new
articles in MEDLINE
about “cell cycle”
Citations
6000
5000
4000
3000
2000
Leland Hartwell
Tim Hunt
Sir Paul Nurse
1000
0
1993 1994 1995
1996 1997
1998 1999 2000
2001 2002 2003
Papers about microarray analysis of the
cell cycle have been extensively cited
Spellman et al., 1998 (yeast)
697 citations
Cho et al., 1998 (yeast)
385 citations
Cho et al., 2002 (human)
61 citations
"for their discoveries of key
regulators of the cell cycle“
Cell cycle related research topics:
Cancer, apoptosis, development, metabolic
engineering, biomass increase, fertility,
protein degradation, phosphorylation,
signalling, etc.
A complex biological system
Evolutionary Conservation & Check-points
• The core machinery of the cell cycle is largely
conserved throughout eukaryotes (from yeast to
humans)
• Especially the multi-cellular organisms have
extensive and sophisticated control mechanisms
(check-points) that check if one step (e.g. DNA
replication) is completed and error-free, before
proceeding to the next steps (e.g. segregation of
the chromosomes)
Different levels of control in the cell
The activity of a protein depends on
–
–
–
–
–
–
regulation of its transcription (by activator/inhibitors)
its mRNA stability
the efficiency of translation (from mRNA to protein)
stability of the mature protein
the subcellular localization of the protein
regulation of the protein function by binding of activator and
inhibitors
– regulation of the protein function by post-translational
modification (e.g. phosphorylation)
– targeted degradation of the protein (e.g. by ubiquitination)
The “just-in-time” principle
• Some products you use all the
time, while others are only
needed once in a while
• People tend to buy or order
things right before they need
them…
Just-in-time synthesis in the cell cycle
Periodically expressed genes
G1
S
G2
M
G1
S
G2
M
Outline
• Introduction to cell division and the cell cycle
• Cell culture synchronization and arrest methods
• Microarrays and periodically expressed genes
• Our own research
In a normal culture, cells grow out of synchrony
Arrest-and-release synchronization
• Temperature sensitive mutants – unable
to pass a certain point in the cell cycle at
high temperatures
• Alpha factor – all cells arrest in the same
stage until a factor is supplied to the culture
Outline
• Introduction to cell division and the cell cycle
• Cell culture synchronization and arrest methods
• Microarrays and periodically expressed genes
• Our own research
Using microarrays with synchronized cells
Synchrone
yeast culture
•
•
•
•
•
Microarrays
Gene expression
Expression Profile
Take samples from a synchronized cell culture
Hybridize each sample to a microarray chip
Normalize to make the chips comparable
String the data-points together to a profile
Look for genes whose expression is periodic
Time-series versus comparative studies
• Most microarray experiments compare measurements
from condition A with condition B to find those genes
that are differentially expressed
– the typical analysis is a statistical test, e.g. t-test
• Time-series experiments are usually aimed at studying
how gene expression changes over time (i.e. identifying
genes that show a specific pattern of expression)
– you can’t make a t-test, instead you have to look for some
pattern!
Look for sine/cosine-like waves
G1
S
G2
M
G1
S
G2
M
Measuring Periodicity
• Fourier-scoring: extracting the ”signal” at the
interdivision frequency.
Fourier score distribution
The thresholding problem
Yeast
Human
Sensitivity versus Specificity
edible
poisonous
Sensitivity: how many of the edible mushrooms do you accept
Specificity: how many of the mushrooms you accept are edible
Sensitivity versus Specificity
edible
find a lot, of which
most is right
Specificity
find little, of which
most is right
find little, of which
most is wrong
find a lot, of which
most is wrong
Sensitivity
poisonous
Sensitivity: how many of the edible mushrooms do you accept
Specificity: how many of the mushrooms you accept are edible
x
The thresholding problem
• Every threshold corresponds to a sensitivity and a specificity
• There is (almost always) a trade-off between the two
Outline
• Introduction to cell division and the cell cycle
• Cell culture synchronization and arrest methods
• Microarrays and periodically expressed genes
• Our own research
Benchmark of computational methods
Many different computational
methods have been developed
to find periodically expressed
genes, but we were the first to
compare them to see which
ones work the best!
Benchmarking of periodicity analysis methods
Computational Methods
Benchmark sets
• Cho et al., 1998
B1 113 known periodic genes from small
scale experiments
– visual inspection
• Spellman et al., 1998
– fourier scoring and correlation
• Zhao et al., 2001
– single pulse model
• Johansson et al., 2003
B2 352 genes bound by at least one of nine
known cell cycle transcription factors,
excluding those in set B1
B3 518 genes annotated as “cell cycle and
DNA processing” in MIPS, excluding
“meiosis” and those in set B1
– partial least squares regression
• Luan & Li, 2004
– cubic spline fitting
• Lu et al., 2004
– Bayesian, mixture model
• de Lichtenberg et al., 2004/2005
– permutation based statistics
How do we benchmark?
We assume that the benchmark sets B1-B3
are enriched in cell cycle regulated genes.
We then benchmarks the ability of each
computational method to retrieve these genes
from the data (rank them high)
Comparison of different methods
--- random
Cho et al.
Spellman et al.
Zhao et al.
Johansson et al.
Luan & Li
Lu et al.
de Lichtenberg et al.
Our periodicity analysis of microarray data
“is the gene significantly regulated?”
– p(regulation): compare actual standard deviation to
randomly sampled background distribution
“is the expression pattern periodic?”
– p(periodicity): compare the actual fourier signal to
distribution made by randomly shuffling the data
point within each profile
Combined score with penalty function
– Combi score: Combination of p-values for regulation
and periodicity, with penalty terms that require both
components to be significant.
Regulation versus periodicity
--- random
regulation alone
periodicity alone
combined
combined + re-norm
combined + consistency
Better methods give better overlap
Why not perfect overlap?
• Different experimental
conditions, such as cell culture
synchronization methods
• Handling of samples,
amplification, hybridization, etc.
• DNA microarray data are noisy
and replicates rarely overlap
completely.
• Signal-to-noise ratio low for
weakly expressed –and
regulated genes
Finding the weakly expressed genes
List of putative cell
cycle regulated genes
Validation of the predictions
• Used a temperature sensitive mutant yeast strain
to synchronize a cell culture in a batch fermentor
• Took out samples every 20 min
• Treated and froze the samples
• Extracted RNA from the samples
• hybridized to custom made DNA microarrays,
where the probes were designed with our own
software tool, OligoWiz
• Performed the raw data treatment and
normalization
• Analyzed the data with our own permutationbased method, which is good at finding weakly
expressed –or regulated genes.
Experimental setup - sampling
– Samples were shaken with
ice (2°C I a few sec)
– Centrifuged 1-1/2 min.
– Supernatant discarded
– Frozen in liquid N2 (-196°C)
– Stored at -80°C until RNA
extraction
Monitoring synchrony
The culture exhibited synchrony with an interdivision time of ~148
minutes, meaning that the experiment covers two entire cell cycles
The results
The results
New genes are weakly expressed
Examples of new discoveries
Mapping gene expression on networks
Node selection
List of periodically
expressed proteins with
peak time
Expand the set of proteins to include
non-periodic proteins that are
strongly connected to periodic
proteins.
Raw Data
Interactions
cell cycle
microarray
data
Physical PPI
interactions with
confidence scores
Require compatible
compartments and high
confidence
Extract cell cycle network
Dynamics of protein complex formation
Thanks to all the other people involved!
CBS, BioCentrum-DTU, Denmark
Søren Brunak (Professor, Center director)
Steen Knudsen (Professor)
Rasmus Wernersson (synchronization, sampling, probe/array design)
Thomas Skøt Jensen (design & data analysis)
Henrik Bjørn Nielsen (array normalization & data analysis)
Anders Fausbøll (data analysis)
Flemming Bryde Hansen (sampling)
EMBL, Heidelberg, Germany
Lars Juhl Jensen (Prediction and benchmarking)
Peer Bork (group leader)
http://www.cbs.dtu.dk/cellcycle
Further reading…
New Weakly Expressed Cell Cycle Regulated Genes in Yeast
Ulrik de Lichtenberg, Rasmus Wernersson, Thomas Skøt Jensen, Henrik Bjørn Nielsen, Anders Fausbøll,
Peer Schmidt, Flemming Bryde Hansen, Steen Knudsen and Søren Brunak
Submitted, 2005
Dynamic Complex Formation During the Yeast Cell Cycle
Ulrik de Lichtenberg, Lars Juhl Jensen, Søren Brunak and Peer Bork
Science, 307 (5710), 724-727, 2005 [PubMed]
Comparison of Computational Methods for the Identification of Cell Cycle Regulated Genes
Ulrik de Lichtenberg, Lars Juhl Jensen, Anders Fausbøll, Thomas Skøt Jensen, Peer Bork and Søren Brunak
Bioinformatics, 2004 (advance online access) [PubMed]
Protein Feature Based Identification of Cell Cycle Regulated Proteins in Yeast
Ulrik de Lichtenberg, Thomas Skøt Jensen, Lars Juhl Jensen and Søren Brunak
Journal of Molecular Biology, 239(4), 663-674, 2003 [PubMed]
http://www.cbs.dtu.dk/cellcycle