From: cyly on
I am now doing a project about cancer detection in matlab.
I've got 120 samples which have classified as tumor or normal cells from patients.
For each sample, there is 19200 parameters which are gene information of the cell.
My objective is to find the best combination of parameters to classify the sample whether it is tumor or normal.

I have already use t-test in matlab to pick out about five hundreds of parameters that is significant. But I want to further reduce the number of parameters to below one hundred. How can I do this in matlab ? thanks very much
From: Walter Roberson on
cyly wrote:
> I am now doing a project about cancer detection in matlab.
> I've got 120 samples which have classified as tumor or normal cells from patients.
> For each sample, there is 19200 parameters which are gene information of the cell.
> My objective is to find the best combination of parameters to classify the sample whether it is tumor or normal.
>
> I have already use t-test in matlab to pick out about five hundreds of parameters that is significant. But I want to further reduce the number of parameters to below one hundred. How can I do this in matlab ? thanks very much

I suggest you read some of the papers produced by National Research
Council of Canada's Institute for Biodiagnostics, especially those on
which R. L. Somorjai is an author.

I am relatively certain that NRC IBD has done Matlab implementations.
From: cyly on
Thx for your suggestion
However how can i get the related paper from NRC IBD?
just surfing in google?
From: us on
cyly <ynlee519(a)hotmail.com> wrote in message <1809595558.10969.1278848437999.JavaMail.root(a)gallium.mathforum.org>...
> Thx for your suggestion
> However how can i get the related paper from NRC IBD?
> just surfing in google?

well... smart thought - really...

us
From: cyly on
How about using anfis in matlab?
I know that this may help me
I have already read the manual in matlab about anfis
but I don't know how can i start