how many ways do you know to visualise such a affymetrix microarray data ?

Question

[Mo]

Hello,

One of the first question I always have is how to visualise my data before even start analysing it. There are quite some statistical techniques which you might use to decompose your matrix and then for example only visualise linear combination of the entire matrix. However, what if I don't do such techniques to reduce the dimension of the data?

It might look very simple in some way, but believe me, it is very difficult when we need to understand the data in an unsupervised way before start working on a given data,

I uploaded an example data set in the attachment.

Any suggestion?

[Manvendra Singh]

depends on what are the questions you want to ask from your data.

most of significant answers are only when you have normalized your data on quantiles or gcrma or rma.

Answer 1

[andrew]

Hi Mohammad,

It's not clear what tools you are using to analyze your data, but our application, iPathwayGuide, has the ability for users to upload their Affy CEL files (for the most popular arrays for Human, Mouse, and Rat) and get gene expression analysis.

Currently we are doing the QC and normalization in the background, but we will soon offer some visualizations on these steps.

One of the keys is to iterate when removing arrays during the QC steps so as to not remove several arrays at once as you could inadvertently remove an array that would otherwise be included if done iteratively. It's also important to have an objective method for inclusion/exclusion. Many of the current QC methods rely on the user to make a qualitative judgement based on box plot or similar visualization. We use a method that scores each sample. Soon we will have the plot showing why it was included/excluded.

Then during the normalization phase it is important to have a method that will normalize the data and provide some visualizations on this step. We have 4 separate methods we are evaluating, but gcrma seems to work really well and is quite fast. We will be providing the visualizations for this step as well in our next release (March 2015).

By the way, if you haven't tried iPathwayGuide, you should. It's completely free to try. Use it as much as you want. Each analysis comes with a 72-hour preview. Upload and run as many dataset as you want. Have fun.