Bioinformatics

1
Extraction of functional information from
large-scale gene expression data

• Bioinformatics
• 91.580 2003 Spring
• Jianping Zhou

2
Contents

• A prominence feature of cell cycle-regulated
  genes
• ----- show more remarkable and active functions
  than others
• SP( shortest-path) analysis to extract
  functional information
• ----- An alternative and complementary to
  clustering analysis

3
Prominence Feature

• Assumption

• Because of their ruling features, the cell
  cycle-regulated genes are assumed to be more
  active and remarkable than others in the Yeast
  Saccharomyces cerevisiae genome.
• When performing filtering process against
  original dataset by some thresholds in terms of
  significance, if the cell cycle-regulated genes
  show higher survival ratio than others, we may
  conclude they are more active and remarkable

4
Prominence Feature

• Methods

• The preprocess utility of Gepas package can be
  used to prepare the comparing dataset
• Microarray gene data are the ideal data sources
• 800 Spellmans identified cell cycle-regulated
  genes for Yeast Saccharomyces cerevisiae are the
  most complete spectrum at this point

5
Prominence Feature

• Methods (cont)

Use a single sentence of Common Lisp to count the
hitting genes (length (intersection regu '(plain
text file content)))  regu the preset CL list
representing the list of 800 cell cycle-regulated
gene names. It is defined in CL as (setf regu
(plain text of 800 cell cycle-regulated gene))
The plain text of 800 cell cycle-regulated gene
can be got by copy and paste of ORF column of
CellCycle98.xls  plain text file content Copy
and paste of preprocess or clustering output
plain text file inside which the ORFs
corresponding to selected genes are contained.
6
Prominence Feature

• Steps

7
Prominence Feature

• Steps (cont)

8
Prominence Feature
 

• Steps (cont)

 
9
Prominence Feature

• Steps (cont)

Parameter Pe, Pk, Sd Pe Minimum percentage of
existing values -- patterns with missing values
greater this rate will be removed. Pk Minimum
number of peaks -- patterns with peak values less
this value will be removed. Sd Threshold for
standard deviation -- patterns with a standard
deviation below the threshold will be
removed. P0 total profiles in the original
file P1 Removed profiles with missing values,
determined by Minimum percentage of existing
values P2 Profiles mended through imputing
missing values, determined by Minimum number of
peaks P3 Removed profiles through filtering out
flat profiles by number of peaks P4 Removed
profiles through filtering out flat profiles by
standard deviation P5 Profiles remaining in the
result dataset Hit Count of genes existing in
both result dataset and 800 Spellman cell
cycle-regulated gene dataset. Hit rate Hit / P5
10
Prominence Feature

• Result

11
Prominence Feature

• Result (cont)

Pe 95
12
SP( shortest-path) analysis

• Introduction

• SP( shortest-path) analysis is used to identify
  transitive genes between two given genes from the
  same biological process.
• Transitive expression similarity among genes can
  be used as an important attribute to link genes
  of the same biological pathway.
• Recent advances in computational and
  experimental technologies have opened up real
  opportunities for annotating gene functions not
  only at the phenomenological levels but also at
  the mechanistic levels.

13
SP( shortest-path) analysis

• Discovery

• With Yeast Saccharomyces cerevisiae genome, The
  author, X. Zhou 5, constructed the cytoplasm
  graph (another two graphs include mitochondria,
  nucleus), which contain 398 genes. All those
  genes are got involved in the same biological
  pathway.
• Through matching the cytoplasm outcome with
  Spellman CellCycle98.xls, six genes are
  identified, they are
• YPR045C YPL221W(BOP1) YIL056W YHR029C YDR130C
  YBR053C

14
SP( shortest-path) analysis

• Discovery (cont)

• Referring to CellCycle98.xls, all these genes
  are with unknown process and far away cluster
  order number each other.
• For the SOM clustering output with respect to
  normalized file, which has 561 hits with 800
  Spellman genes, those genes exist in YPR045C
  Cluster (2, 4) YPL221W Cluster (1, 1) YBR053C
  Cluster (2, 7). Other three are not found.
• As far as all my clustering outputs, none is
  found in clustering.
• All Ftigo linked databases have no results for
  these five genes or ORFs
• No evidence show these six genes can stay in the
  same cluster.

15
References
1 Paul T. Spellman, Gavin Sherlock,Michael Q.
Zhang, Vishwanath R. Iyer, Kirk Anders, Michael
B. Eisen, Patrick O. Brown, David Botstein, and
Bruce Futcher Comprehensive Identification of
Cell Cycle-regulated Genes of the Yeast
Saccharomyces cerevisiae by Microarray
Hybridization MBC, Vol. 9, Issue 12, 3273-3297,
December 1998 2 Oliveros, J.C., Blaschke, C.,
Herrero, J., Dopazo, J. Valencia, A. (2000)
Expression profiles and biological function.
Genome Informatics Workshop 2000, 11, 106-117 3
M. Q. Zhang Extracting functional information
from microarrays A challenge for functional
genomics PNAS, October 1, 2002 99(20) 12509 -
12511. 4 M. Q. Zhang Large-Scale Gene
Expression Data Analysis A New Challenge to
Computational Biologists Genome Res.,
August 1, 1999 9(8) 681 - 688. 5 X. Zhou,
M.-C. J. Kao, and W. H. Wong From the Cover
Transitive functional annotation by shortest-path
analysis of gene expression data PNAS,
October 1, 2002 99(20) 12783 - 12788. 6
www.biostat.harvard.edu/complab/SP/