|Dataset: Effect of Plasmodium falciparum infected erythrocytes on primary human brain microvascular endothelial cell (GEO: GSE9861)|
Study: investigated the global transcriptional gene response of primary human brain endothelial cells after incubation with high numbers of infected erythrocytes.
Chip: (GPL571) Affymetrix GeneChip Human Genome U133A 2.0 Array
Samples: Total of 8 samples (4 control and 4 treated) were analyzed. 4 control samples included two normal RBC control and two medium controls. 4 treated samples includes 2 exposed to low binding Pf-IRBC and 2 exposed to high binding Pf-IRBC (Pf-IRBC-P). Medium and RBC controls were finally used as four replicates of control and all four Pf-IRBC or Pf-IRBC-P exposed endothelial cells were used as 4 separate treated controls.
Normalized: Signal intensity was calculated by GCOS 1.4 software - but affy values are NOT NORMALIZED!!
Karsten performed normalization - email:
I have placed files with normalized values into /tmp on koch:
khokamp@koch:/tmp> l GSE9861_*
-rw-r--r-- 1 khokamp wg-users 3257047 2009-05-13 13:56 GSE9861_clean_gcrma.txt
-rw-r--r-- 1 khokamp wg-users 3256320 2009-05-13 13:56 GSE9861_gcrma.txt
But if you are going to analyse them with limma, you could easily re-do
this yourself in R:
1. Change into the directory containing the (compressed) CEL files, e.g.
2. start R
3. read data and normalize:
data = ReadAffy()
norm_data = gcrma(data)
norm_data is an expressionSet that should be readily usable with limma
The last step that I applied for extracting the data was
It might be best to run both the original and the clean data set through
limma to see if the resulting gene lists differ greatly.
Explanation of 'clean' version:
this experiment has been loaded into ArrayWiki...http://arraywiki.bme.gatech.edu/index.php/GSE9861
As you can see, the first slide has some fairly strong artifacts and gets the lowest quality score (88.63%).
They provide a 'clean' version of the original data, where the spots with high variance have been replaced with
the median value of this probe from other chips in the dataset (see here for more details: