omiras: MicroRNA regulation of gene expression Sören Müller, Goethe University of Frankfurt am Main Molecular Bioinformatics Group, Institute of Computer Science Plant Molecular Biology Group, Institute of Molecular BioSciences
Regulation of gene expression by micrornas Winter, Julia, et al. "Many roads to maturity: microrna biogenesis pathways and their regulation." Nature cell biology 11.3 (2009): 228-234.
Exponential growth of mirna publications Kozomara, A., & Griffiths-Jones, S. (2011). mirbase: integrating microrna annotation and deep-sequencing data. Nucleic acids research, 39(suppl 1), D152-D157.
What do we learn about functions? -MiRBase v20 contains ~1800 primary mirna transcripts for human -One mirna can potentially control more than 100 genes -Currently only ~ 5000 validated interactions present in databases, ~2000 of these for human -Reason: Wet-lab work very time consuming (transfection, luciferase report assays...) -Sequencing technologies (e.g. PARCLIP) are still under development and not widely used -Interactions might even be cell/tissue specific Muniategui, Ander, et al. "Joint analysis of mirna and mrna expression data." Briefings in bioinformatics 14.3 (2013): 263-278.
What is the accuracy of predictions? - The estimated false positive rate of predictions ranges from 24% to 70% lack of concordance between the predictions of different databases Muniategui, Ander, et al. "Joint analysis of mirna and mrna expression data." Briefings in bioinformatics 14.3 (2013):263278.
How to minimize wet-lab work Grab the low hanging fruit! Focus on: -A quite simplistic experimental design for sequencing (two-condition comparison) -Interactions predicted by several tools
OmiRas for the differential expression analysis of srna-seq experiments omiras can: 1. Control the quality of your data 2. Check the expression of ncrna in each library 3. Test for differential expression between conditions 4. Visualize interactions between genes and mirnas predicted by several prediction tools as well as validated interactions omiras needs: -Raw srna-seq data in FASTQ/FASTA format from two conditions with or without replicates omiras is available at: http://tools.genxpro.net/omiras/ For details see: Müller, Rycak, et al. "omiras: A Web server for differential expression analysis of mirnas derived from small RNA-Seq data " Bioinformatics 2013, btt457v1-btt457.
The omiras workflow
Time consumption Processing of srna libraries from Osanto et. al (2012) (Available at GEO: GSE37616). -13 libraries of clear cell renal cell carcinoma and adjacent tissue -Library size between 1 and 3 million smallrna-seq reads -Processing time ~ 2 hours ~9 minutes for each library -Time needed for bowtie mapping: ~7 minutes
An example: Small RNA-Sequencing of pancreatic cancer and adjacent control tissue
Low 5-year survival rate in PDAC
Result overview
Quality control
Sequencing quality 22.08.13 14
Gel extraction 22.08.13 15
Different degrees of rrna contamination M1052_Normal 22.08.13 M1040_Normal 16
Different degrees of rrna contamination M1052_Normal 22.08.13 M1040_Normal 17
Differential expression
Library comparison: Scatter-/Correlation plots 22.08.13 19
Differential Expression 22.08.13 20
Functional assignment
MiRNA-mRNA interaction network: Input 22.08.13 22
MiRNA-mRNA interaction network...the regulation of FN1 by mir-217 is one of the hallmarks of pancreatic cancer development...the mir-217 microrna functions as a potential tumor suppressor in pancreatic ductal adenocarcinoma by targeting KRAS mir-216b suppresses tumor growth and invasion by targeting KRAS mir-217 controls the osteogenic activity of Runx2...THBS1 is strongly expressed in most pancreatic adenocarcinomas and is likely to contribute to the extensive neovascularization and spread of this highly aggressive tumor....therefore, ADAM9 overexpression might contribute to the aggressiveness of PDACs....Rac1 expression increased in mouse and human pancreatic tumors, particularly in the stroma. Deletion of Rac1 in KrasG12D-induced PDAC in mice reduced formation of ADM, PanIN, and tumors and significantly prolonged survival....patients with prostate cancer exhibiting immunohistochemically detectable nuclear E2F3 expression have poorer overall survival (P=0.0022) and cause-specific survival (P=0.0047) than patients without detectable E2F3 expression...nrp2 expression was found in pancreatic cancers and directly promotes tumor progression in a cell autonomous manner 22.08.13 23
Loss of mir-29c leads to increased expression of genes coding for ECM proteins 22.08.13 24
Outlook -Knockdown/Transfection experiments for mirnas of interest -Develop a method to predict effects of a microrna knockdown/transfection on the network derived from static data in combination with information from several databases 22.08.13 25
22.08.13 Thank you for your attention! 26
Main assumption of prediction algorithms: Conservedness Source: http://www.targetscan.org