Protoarray CSF Analysis

Preprocess

Positive control: AlexaAntiMouseAb~N/A, HumanIgG1~N/A, HumanIgG2~N/A, HumanIgG3~N/A, HumanIgG4~N/A, Anti-HumanIgG1~N/A, Anti-HumanIgG2~N/A, Anti-HumanIgG3~N/A, Anti-HumanIgG4~N/A, V5control5~N/A, V5control4~N/A, V5control3~N/A, V5control2~N/A, V5control1~N/A

Negative control: Buffer, BSA~N/A, GST1~RFU:322.74, GST2~RFU:578.54, GST3~RFU:1770.2, GST4~RFU:4230.18

Preprocessing involves down-weighing the positive controls and up-weighing the negative controls. Positive controls should be ubiquotously expressed while we should expect little or no expression at the negative control spots.

Read ../rawdata/control_67336_18052013.gpr 
Read ../rawdata/control_68051_21052013.gpr 
Read ../rawdata/control_68213_11052013.gpr 
Read ../rawdata/control_68473_05062013.gpr 
Read ../rawdata/Grade_L_68204_11052013.gpr 
Read ../rawdata/Grade_L_67341_18052013.gpr 
Read ../rawdata/Grade_L_68210_13052013.gpr 
Read ../rawdata/Grade_III_67346_18052013.gpr 
Read ../rawdata/Grade_III_68001_21052013.gpr 
Read ../rawdata/Grade_III_68052_21052013.gpr 
Read ../rawdata/Grade_III_68207_11052013.gpr 
Read ../rawdata/Grade_III_68211_11052013.gpr 
Read ../rawdata/Grade_III_68476_18052013_2.gpr 
Read ../rawdata/Grade_IV_68037_21052013.gpr 
Read ../rawdata/Grade_IV_68056_21052013_2.gpr 
Read ../rawdata/Grade_IV_68209_11052013.gpr 
Read ../rawdata/Grade_IV_68212_11052013.gpr 
Read ../rawdata/Grade_IV_68481_18052013.gpr 
Read ../rawdata/Grade_IV_68055_21052013_2.gpr 
Read ../rawdata/Grade_IV_67347_18092013_2.gpr 

Intensity distribution pre normalization

Internsity distribution post normalization

PCA pre normalization

Performing principle component analysis on raw intensities, which have not been background corrected yields no separation for Low or high grade samples from the controls.

PCA post normalization

The variance explained here is only around 7.26% by the first principal component. Since the PCA plots do not show a lot of separation, we expect a handful of genes to be differentially expressed.

Heatmap post normalization

The samples are all over, with no well-defined clusters.

MDS plot post normalization

Multidimensional Scaling plot gives an overview of the level of similarity of samples. The points in the plot are placed such that their distances in the higher dimensions (if we have 10000 genes, we have 10000 dimensions) are preserved.

We do see that Grade4 samples seem to cluster together hinting that they are less heterogenous. While there is lot of heteregenity in the Grade3 samples.

Differential Expression

Since the sample size is too less, we do not remove any samples even though they look like outliers on the MDS plot.