Guided Tutorial:

Using GSEA as an analytical tool for molecular profiling

Getting Started: Create a GSEA workstation that works for you

Desktop/Laptop  running  Mac  OS  X,  Windows,  or  Linux:    

 Java  6  or  7  (Desktop  GUI  or  command  line)    

 R    

 GenePa4ern  (Web,  Desktop,  or  Server)      

When  would  you  use  GSEA?    What  do  you  need  to  get  started?    

 Basic  workstaIon  with  Java    Expression  files  converted  to  .gct  (GenePaNern)    Phenotype  labels  (any  text  editor)    Gene  Sets  (public  repository  or  custom  sets)  

 

Getting Started: Working with GSEA files

1.  Sources  of  gene  expression  data  (.gct  files):    

 Microarrays  (.cel  files)    

 RNA-­‐Seq  (.fpkm_tracking  file)    2.  Phenotype  labels  (.cls  files)    3.  Gene  Sets  (MSigDB  or  custom):    

 PosiIonal    

 Curated    

 Pathway  (KEGG,  Biocarta,  etc.)  

Getting Started: GCT file column structure

#1.2  23230  2  gene_short_name  DescripIon        Untreated_FPKM  CisplaIn_FPKM  Mcm4      chr16:15623989-­‐15637493  16.5908      33.9435  Mcm5      chr8:77633426-­‐77652338  5.97553      16.5161  Mcm6      chr1:130228167-­‐130256233  12.7255      28.2007  Mcm7      chr5:138605816-­‐138613090  25.7281      49.4802  

Getting Started: Converting from microarray .cel expression to .gct file

Getting Started: Converting from RNA-Seq expression genes.fpkm_tracking to .gct file

Getting Started: Real World Example

GCT  file  structure:    #1.2  23230  2  gene_short_name  DescripIon  Untreated    CisplaIn  GeneA      na      16.5908      33.9435  GeneB      na      5.97553      16.5161  GeneC      na      12.7255      28.2007  GeneD      na      25.7281      49.4802  

Getting Started: Creating a .cls file to label phenotypes

Getting Started: Real World Example

CLS  file  structure:    2  2  1  #  Untreated  CisplaIn  Untreated  CisplaIn  

GCT  file  structure:    #1.2  23230  2  gene_short_name  DescripIon  Untreated    CisplaIn  GeneA      na      16.5908      33.9435  GeneB      na      5.97553      16.5161  GeneC      na      12.7255      28.2007  GeneD      na      25.7281      49.4802  

Running GSEA: Analyzing cancer cell lines for p53 targets

P53_hgu95av2  (50  p53  MUT  or  WT  cell  lines  analyzed  on  the  HG_U95Av2  (Affy)  Human  Genome  U95  GeneChip:    #1.2  12625  50                              NAME  DescripIon  786-­‐0  BT-­‐549  …  UACC-­‐62    UO-­‐31  100_g_at  na      215.37  132.94  …  451.01    186.68    1000_at  na      328.68  234.31  …  381.41    285.72    1001_at  na      39.64  8.84    …  40.52    31.88      P53.cls  (MUT  vs.  WT):  50  2  1  #MUT  WT  MUT  MUT  …  WT  WT    Chip:  HG_U95Av2.chip  

Download  and  load  the  p53  datasets  from  the  GSEA  website:  hNp://www.broadinsItute.org/gsea/datasets.jsp  

Running GSEA: Analyzing cancer cell lines for p53 targets

Running GSEA: Analyzing cancer cell lines for p53 targets

GSEA Reports: Analyzing cancer cell lines for p53 targets

GSEA Reports: Analyzing cancer cell lines for p53 targets

GSEA Reports: Analyzing cancer cell lines for p53 targets

Leading Edge Analysis: Analyzing cancer cell lines for p53 targets

Leading Edge Analysis: Analyzing cancer cell lines for p53 targets

Leading Edge Analysis: Analyzing cancer cell lines for p53 targets

Browsing MSigDB: Analyzing cancer cell lines for p53 targets

References

GSEA:  hNp://www.broadinsItute.org/gsea/    GenepaNern:  hNp://www.broadinsItute.org/cancer/sojware/genepaNern/  

ARACNE demo

Xi Zhao, PhDCancer Center for Systems Biology, Stanford University

xi.zhao@stanford.edu

• ARACNE theory recap

• Java GUI

• Command line tool

Friday, May 10, 13

Recap: interactome

• Genes do not function alone

• Construct gene interaction network from genomic data

Elements of Statistical Learning (2nd Ed.) c©Hastie, Tibshirani & Friedman 2009 Chap 1

SID42354SID31984SID301902SIDW128368SID375990SID360097SIDW325120ESTsChr.10SIDW365099SID377133SID381508SIDW308182SID380265SIDW321925ESTsChr.15SIDW362471SIDW417270SIDW298052SID381079SIDW428642TUPLE1TUP1ERLUMENSIDW416621SID43609ESTsSID52979SIDW357197SIDW366311ESTsSMALLNUCSIDW486740ESTsSID297905SID485148SID284853ESTsChr.15SID200394SIDW322806ESTsChr.2SIDW257915SID46536SIDW488221ESTsChr.5SID280066SIDW376394ESTsChr.15SIDW321854WASWiskottHYPOTHETICSIDW376776SIDW205716SID239012SIDW203464HLACLASSISIDW510534SIDW279664SIDW201620SID297117SID377419SID114241ESTsCh31SIDW376928SIDW310141SIDW298203PTPRCSID289414SID127504ESTsChr.3SID305167SID488017SIDW296310ESTsChr.6SID47116MITOCHONDRChrSIDW376586HomosapiensSIDW487261SIDW470459SID167117SIDW31489SID375812DNAPOLYMESID377451ESTsChr.1MYBPROTOSID471915ESTsSIDW469884HumanmRNASIDW377402ESTsSID207172RASGTPASESID325394H.sapiensmRNGNALSID73161SIDW380102SIDW299104

BREAST

RENAL

MELANOMA

MELANOMA

MCF7D-repro

COLON

COLON

K562B-repro

COLON

NSCLC

LEUKEMIA

RENAL

MELANOMA

BREAST

CNS

CNS

RENAL

MCF7A-repro

NSCLC

K562A-repro

COLON

CNS

NSCLC

NSCLC

LEUKEMIA

CNS

OVARIAN

BREAST

LEUKEMIA

MELANOMA

MELANOMA

OVARIAN

OVARIAN

NSCLC

RENAL

BREAST

MELANOMA

OVARIAN

OVARIAN

NSCLC

RENAL

BREAST

MELANOMA

LEUKEMIA

COLON

BREAST

LEUKEMIA

COLON

CNS

MELANOMA

NSCLC

PROSTATE

NSCLC

RENAL

RENAL

NSCLC

RENAL

LEUKEMIA

OVARIAN

PROSTATE

COLON

BREAST

RENAL

UNKNOWN

FIGURE 1.3. DNA microarray data: expression ma-

sample (N)

gene

(p)

e.g Expression matrix

gene

(p)

gene (p)

Pair-wise correlation matrix

ARACNE: Mutual Information

Basso et al 2005

ARACNE: Data Processing Inequality

Post processing: e.g. correlation significance

Friday, May 10, 13

Recap: ARACNE

• Mutual Information (MI):

• detect non-linear dependencies between a pair of variables (X, Y)

• Data Processing Inequality (DPI):

• remove weakest link (tends to be indirect interaction) in a triple

TF

nTF I1 I2

I3 nTF

G1

G2 G3

I1 > I2 > I3

TF

nTF I1 I2

I3 nTF

G1

G2 G3

Friday, May 10, 13

ARACNE limitations1. MI has low statistical power although detect non-linear

dependencies

MI “has lower power than distant correlation, ... (it) is sometimes less powerful than Pearson correlation as well, the linear case being particularly worrisome.” (Simon and Tibshirani, 2011 comment on Reshef et al 2011)

2. Can easily turn into Hair balls: Difficulty in interpretation

Martin Krzywinski

Friday, May 10, 13

ARACNE java GUI: Setup• Download (http://wiki.c2b2.columbia.edu/califanolab/index.php/Software/ARACNE)

• aracne.zip (http://wiki.c2b2.columbia.edu/califanolab/download/ARACNE/aracne.zip): • Java JDK 1.5 or newer (compiling issues with 1.7 on Mac?)

• Set $PATH variable (on Mac)

• edit .bashrc file (e.g vi ~/.bashrc)

• edit launch_aracne.sh

• Launch from terminal:• $cd <aracne folder>

• $sh launch_aracne.sh

Friday, May 10, 13

ARACNE file requirement• Input file - expression matrix (.exp):

• TAB-delimited text files

• Output file - adjacency matrix file (.adj):

• TAB-delimited text files

optional

Friday, May 10, 13

ARACNE java GUI• Load “BCell_matrix.exp”

• Construct network around hubs (optional)

• TF list (optional)

• Name the output file (optional)

• Filter by p value (optional)

• Filter by DPI tolerance (optional): 0~0.2

• Output: adjacency file “.adj”

• Load existing adjacency file “.adj”(optional)

Friday, May 10, 13

ARACNE java GUI• Network visualization

• Network Browser -> Load -> Draw -> Cytoscape

Friday, May 10, 13

ARACNE Command Line

$cd <ARACNE.src/ARACNE folder>

Construct network using all probes provided in BCell_matrix.exp

$./aracne2 -i ../../aracne/Data/BCell_matrix.exp

Construct network around probes provided in myc_probes.txt

$./aracne2 -i ../../aracne/Data/BCell_matrix.exp -s ../../aracne/Data/myc_probes.txt

Load existing adjacency file .adj with DPI & P-value filtering

$./aracne2 -i ../../aracne/Data/BCell_matrix.exp -j ../../aracne/Data/BCell_matrix_k0.139.adj -p 0.001 –e 0.15

Download ARACNE.src.tar.gzARACNE options: http://wiki.c2b2.columbia.edu/califanolab/images/9/92/Usage.txt-i <file> Input gene expression profile dataset-o <file> Output file name (optional)-j <file> Existing adjacency matrix (.adj) file-e <tolerance> DPI tolerance-p <p-value> P-value for MI threshold (e.g. 1e-7)

Friday, May 10, 13

References• Basso, K., Margolin, A. a, Stolovitzky, G., Klein, U., Dalla-Favera, R., & Califano, A. (2005). Reverse

engineering of regulatory networks in human B cells. Nature genetics, 37(4), 382-90. doi:10.1038/ng1532

• Margolin, A. a, Nemenman, I., Basso, K., Wiggins, C., Stolovitzky, G., Dalla Favera, R., & Califano, A. (2006). ARACNE: an algorithm for the reconstruction of gene regulatory networks in a mammalian cellular context. BMC bioinformatics, 7 Suppl 1, S7. doi:10.1186/1471-2105-7-S1-S7

• Margolin, A. a, Wang, K., Lim, W. K., Kustagi, M., Nemenman, I., & Califano, A. (2006). Reverse engineering cellular networks. Nature protocols, 1(2), 662-71. doi:10.1038/nprot.2006.106

• Carro, M. S., Lim, W. K., Alvarez, M. J., Bollo, R. J., Zhao, X., Snyder, E. Y., Sulman, E. P., et al. (2010). The transcriptional network for mesenchymal transformation of brain tumours. Nature, 463(7279), 318-25. Nature Publishing Group. doi:10.1038/nature08712

• SIMON, N., and TIBSHIRANI R.. "comment on “detecting novel associations in large data sets” by Reshef et al, Science dec 16, 2011." Science (2011).

• Reshef et al (2011) Detecting Novel Associations in Large Data Sets. Science: 1518-1524. [DOI:10.1126/science.1205438]

Friday, May 10, 13