Simulated GEO matrix¶
Mutations in DNA that is coding for proteins (genes) or regulatory elements are often the cause of most diseases. Affymatirx arrays allow quantification of specific sequences of DNA or RNA in biological samples such as blood, tissue, or tumors. These arrays give the relative levels of specific transcript (between two samples) so that these can be compared (between eg. healty vs diseased tissues or before vs after treatment).
To do so, DNA sequences unique to specific genes/mutations are printed as small spots on a glass surface (arrays), with each spot containing thousends of copies of a single sequence. The DNA from each sample is than tagged with different colors of flourecent molecules (red or green) and hybredized with the array. This way, DNA from sample can bind to sequences on the surface that are antisense of its own sequence (specific binding). The array undergoes several washing steps to remove the none-specifically bound DNA sequences, which bind loosly. The plate is than scanned and the flourecence intensity for each color at each spot is recorded. This intensity is an indicator of the ammount of DNA bound and probably the levels of that specific DNA present in the sample. The information regarding the specific sequence printed on each spot, control spots, and other array specific information is stored in a CDF format/file. By comparing the intensity of red to green at each spot you can know which sample contained more of that specific transcript.
For this workflow a simulated dataset was used containing patient meta-information, gene expression levels (microarray), and gene ontology information for 500 patients and 500 genes. This workflow performs Standard statistical methods such as linear regression model (lm.fit from stats package), covariance (‘cov’ from ‘stats’ package), biclustering (‘biclust’ package), SVM (‘irlba’ package), and differential gene expression using two sample Wilcoxon test (also know as ‘Mann-Whitney’ test) are performed.
![digraph SIMULATED_GEO_MATRIX {
fontname="sans-serif";
penwidth="0.1";
edge [comment="Wildcard edge",
fontname="sans-serif",
fontsize=10,
colorscheme="blues3",
color=2,
fontcolor=3];
node [fontname="serif",
fontsize=13,
fillcolor="1",
colorscheme="blues4",
color="2",
fontcolor="4",
style="filled"];
subgraph cluster0 {
label="Load data";
edge [comment="Wildcard node added automatic in EG."];
node [comment="Wildcard node added automatic in EG."];
c0_geo [shape="invhouse",
label=<GEO Expression<BR/><I>Simulated, 500x500</I>>];
c0_gen [shape="invhouse",
label=<Gene meta info<BR/><I>Simulated, 500x500</I>>];
c0_pat [shape="invhouse",
label=<Patient meta info<BR/><I>Simulated, 500x500</I>>];
c0_gos [shape="invhouse",
label=<Gene Ontology info<BR/><I>Simulated, 500x500</I>>];
c0_rds [label=<Read and store<BR/><I>readRDS(), list()</I>>];
c0_dat [label=<Data>];
c0_geo -> c0_rds;
c0_gen -> c0_rds;
c0_pat -> c0_rds;
c0_gos -> c0_rds;
c0_rds -> c0_dat;
}
subgraph cluster1 {
label="Perform linear regression analysis";
edge [comment="Wildcard node added automatic in EG."];
node [comment="Wildcard node added automatic in EG."];
c1_dat [label="Data"];
c0_dat -> c1_dat [ltail=cluster0,
lhead=cluster1];
c1_gen [label=<Subset Genes<BR/>based on <I>function</I>>];
c1_res [label=<Select <I>drug.response</I>>];
c1_aaa [label=<Create expression matrix<BR/>for each gene in each <BR/>patient using <I>merge()</I>>];
c1_cst [label=<cast to matrix<BR/><I>patientid x geneid</I>>];
c1_lmf [label=<fit linear regression<BR/>using <I>lm.fit()</I>>];
c1_prt [label=<<I>print()</I>>];
c1_dat -> c1_gen [label="$gen"];
c1_dat -> c1_res [label="$pat"];
c1_dat -> c1_aaa [label="$geo"];
c1_gen -> c1_aaa;
c1_aaa -> c1_cst;
c1_cst -> c1_lmf [label="x"];
c1_res -> c1_lmf [label="y"];
c1_lmf -> c1_prt;
{
rank=same;
edge [comment="Wildcard node added automatic in EG."];
node [comment="Wildcard node added automatic in EG."];
c1_cst;
c1_res;
}
}
subgraph cluster2 {
label="Calculate covariance";
edge [comment="Wildcard node added automatic in EG."];
node [comment="Wildcard node added automatic in EG."];
c2_dat [label="Data"];
c0_dat -> c2_dat [ltail=cluster0,
lhead=cluster2];
c2_pat [label=<Subset patients<BR/>based on <I>disease</I>>];
c2_aaa [label=<Create expression matrix<BR/>for each gene in each <BR/>patient using <I>merge()</I>>];
c2_cst [label=<cast to matrix<BR/><I>patientid x geneid</I>>];
c2_cov [label=<compute covariance<BR/><I>stats::cov()</I>>];
c2_scv [label=<select top 25 % genes<BR/><I>which( > .75 * max())</I>>];
c2_prt [label=<Print complete cases<BR/><I>complete.cases(), print()</I>>];
c2_dat -> c2_pat [label="$pat"];
c2_dat -> c2_aaa [label="$geo"];
c2_pat -> c2_aaa;
c2_aaa -> c2_cst;
c2_cst -> c2_cov;
c2_cov -> c2_scv;
c2_scv -> c2_prt;
}
subgraph cluster3 {
label="Perform bi-clustering";
edge [comment="Wildcard node added automatic in EG."];
node [comment="Wildcard node added automatic in EG."];
c3_dat [label="Data"];
c3_pat [label=<Subset paients<BR/>based on <I>gender & age</I>>];
c3_aaa [label=<Create expression matrix<BR/>for each gene in each <BR/>patient using <I>merge()</I>>];
c3_cst [label=<cast to matrix<BR/><I>patientid x geneid</I>>];
c3_bic [label=<Perform biclustering<BR/><I>biclust(method=BCssvd, K=5)</I>>];
c3_cdf [label=<Convert to data.frame<BR/><I>biclust::writeclust()</I>>];
c3_prt [label=<Print complete cases<BR/><I>complete.cases(), print()</I>>];
c0_dat -> c3_dat [ltail=cluster0,
lhead=cluster3];
c3_dat -> c3_pat [label="$pat"];
c3_dat -> c3_aaa [label="$geo"];
c3_pat -> c3_aaa;
c3_aaa -> c3_cst;
c3_cst -> c3_bic;
c3_bic -> c3_cdf;
c3_cdf -> c3_prt;
}
subgraph cluster4 {
label="Compute largest singular values";
edge [comment="Wildcard node added automatic in EG."];
node [comment="Wildcard node added automatic in EG."];
c4_dat [label="Data"];
c0_dat -> c4_dat [ltail=cluster0,
lhead=cluster4];
c4_gen [label=<Subset Genes<BR/>based on <I>function</I>>];
c4_aaa [label=<Create expression matrix<BR/>for each gene in each <BR/>patient using <I>merge()</I>>];
c4_cst [label=<cast to matrix<BR/><I>patientid x geneid</I>>];
c4_irl [label=<Perform SVD<BR/><I>irlba(bu=50, nv=50)</I>>];
c4_prt [label=<Print complete cases<BR/><I>complete.cases(), print()</I>>];
c4_dat -> c4_gen [label="$gen"];
c4_dat -> c4_aaa [label="$geo"];
c4_gen -> c4_aaa;
c4_aaa -> c4_cst;
c4_cst -> c4_irl;
c4_irl -> c4_prt;
}
subgraph cluster5 {
label="Statistical test";
edge [comment="Wildcard node added automatic in EG."];
node [comment="Wildcard node added automatic in EG."];
c5_dat [label="Data"];
c0_dat -> c5_dat [ltail=cluster0,
lhead=cluster5];
c5_cst_ge [label=<cast to matrix<BR/><I>Gene ID x Patient ID</I>>];
c5_cst_go [label=<cast to matrix<BR/><I>Gene ID x GO ID</I>>];
c5_st1 [label=<Expression levels<BR/>genes in GO term>];
c5_st2 [label=<Expression levels<BR/>genes NOT in GO term>];
c5_wlc [label=<Perform Mann–Whitney test<BR/><I>lapply(), wilcox.test()</I>>];
c5_sbs [label=<Subset significant GOs<BR/><I>subset(p < 1e-3)</I>>];
c5_prt [label=<Print complete cases<BR/><I>complete.cases(), print()</I>>];
c5_dat -> c5_cst_ge [label="$geo"];
c5_cst_ge -> c5_st1;
c5_cst_ge -> c5_st2;
c5_dat -> c5_cst_go [label="$go"];
c5_cst_go -> c5_st1 [label="1"];
c5_cst_go -> c5_st2 [label="0"];
c5_st1 -> c5_wlc [label="x"];
c5_st2 -> c5_wlc [label="y"];
c5_wlc -> c5_sbs;
c5_sbs -> c5_prt;
}
}](../../_images/graphviz-39fc5aba57d8e9926a5c3809d2e271a337fa73cd.png)
Diagram of Simulated GEO Matrix analysis workflow.
Packages and Dependencies¶
There are 3 packages used in this workflow, which depend on 7 additional packages from CRAN (dependencies)
Used packages:
- CRAN: biclust, s4vd, irlba
Package dependencies:
- CRAN: lattice, colorspace, MASS, flexclust, modeltools, biclust, Matrix
Data¶
This workflow uses simulated data from GenBase data_generator.py with 500 as size of columns and rows. Here you can read about the original GenBase study by Taft R et al., 2014.
License¶
- Copyright (c) 2015 MIT DB Group based on code from GenBase
- Copyright (c) 2015 Hannes Mühleisen based on code from GenBase (fork)
- Copyright (c) 2015 Ieuan Clay based on code from genbench
- Copyright (c) 2015-2016 BeDataDriven B.V. License: GPL version 2 or higher