Computational Discovery of miR-TF Regulatory Modules in Human Genome

1
Computational Discovery of miR-TF Regulatory
Modules in Human Genome
Dang Hung Tran1,3, Kenji Satou1,2, Tu Bao Ho1,
and Tho Hoan Pham1,3
1 Japan Advanced Institute of Science and
Technology 2 Kanazawa University 3 Hanoi National
University of Education
Singapore ? September 9-11, 2009
2
Outline

• Background
• Motivation
• Discover miR-TF modules
• Results and discussion
• Conclusion

3
Gene regulation

• Conventional view by
• Transcription factors (TF)
• Activate gene expression
• Depress gene expression

4
What are microRNAs ?

• RNA world
• Coding RNA
• Non-coding RNA

• Recent concerning
• Small non-coding RNAs

• microRNAs
• Class of small non-coding RNAs
• Bind to messenger RNAs
• Repress gene expression

5
microRNA biogenesis
6
Research motivation

• Gene regulation
• Transcriptional level by transcription factors
  (TF)
• Post-transcriptional level by microRNAs (miRNA)
• How TFs and miRNAs function together?
• Investigate coordinated gene regulation by TFs
  and miRNAs may elucidate their functions.

Chen and Rajewsky Nature Reviews Genetics (2007)
We need to Identify regulatory modules that
consist of miRNAs, Transcription Factor, and
regulated genes (miR-TF module)
7
Related works

• Detection of miRNA regulatory modules
• Tran et al. 2008, Joung et al. 2009, Liu et al.
  2009
• Concentrated on finding the relationship between
  miRNAs and their target genes
• Finding the relationships between TFs and miRNAs
• Cui et al. 2007, Shalgi et al. 2007, Zhou et al.
  2008
• Showed that miRNA mediated regulatory circuits
  are prevalent in the human genome
• But the combined roles of miRNAs and other
  factors in gene regulatory network still remains
  unanswered

8
Problem

• Given
• TFs genes binding information
• miRNAs genes binding information
• Find
• miR-TF modules with biological meaning
• Each module consist of three components TFs,
  miRNAs, and genes regulated by them

9
Algorithm

• Going over all genes i
• For each i, search all subsets of miRNAs M
• For each M, search all subsets of TFs T
• Find gene set G, bind by M and T
• If Ggt1 report modules and mark all the subsets
  of M and T

10
Experiment

• Databases
• miRNAs and TFs regulatory signatures from CRSD
  database
• CRSD integrates six well-known databases (e.g.
  UniGene, TRANSFAC, miRBASE)
• Dataset contains
• 267 miRNAs, 483 TFs, and 1253 regulated genes
• Performing algorithm
• Written in C
• Run on Linux operating system

Schematic description of the data set construction
11
Evaluation and findings (1/4)

• Tested binding data at a number of different
  confidence level (0.05, 0.01, 0.005, 0.001)
• Each module consists of three components
• The number modules found by our algorithm is as
  follows

• For further analysis, we selected the values of
  p1 and p2 equal to 0.01. (182 modules)

12
Evaluation and findings (2/4)

• Some of miR-TF modules share a subset of miRNAs
  or TFs on regulation of the target genes
• For example modules 66, 67, 68
• Three modules share a common miRNA hsa-miR-125b
• Two of them share two miRNAs hsa-miR-125a and
  hsa-miR-125b
• Two of them have the same TF Zic3
• etc.
• The coordinated regulation of target genes by
  miRNAs and TFs is more complicated

Illustration of the three miR-TF modules (66, 67,
68)
13
Evaluation and findings (3/4)

• It is hard to directly validate the miR-TF
  modules
• Using GO to validate modules with respect to
  biological processes
• The set of regulated genes and host genes of TFs
  in modules related to some specific GO terms

14
Evaluation and findings (4/4)

• Found that some miRNAs and TFs in miR-TF modules
  have associations with cancer developments
• Several miRNAs related to lung and other human
  cancer

15
Conclusion

• Proposed a comprehensive search method for
  discovering functional miR-TF modules
• Detected miR-TF modules
• Involved in specific biological processes
  (validation with GO)
• Contain miRNAs and TFs related to several types
  of cancer diseases
• Moreover, provided a view of combinatorial
  regulation of TFs and miRNAs

16
Future work

• Algorithm does not consider the active behavioral
  characteristics of TFs
• Integrate some datasets of gene expression may
  help us to discover more coherent functional
  miR-TF modules

17

• THANK YOU !