Why do we use two channels in MICRO ARRAY analysis ?

[karthik3k]

Why do we use two channels in MICRO ARRAY analysis ??
LIke red and Green ??
What does the ratio between red/green signify ??

 

[Monique]

It are two different samples on the same chip. You label one RNA sample with green fluorescent dye and the other sample with red fluorescent dye. You hybridize that to a chip, if the concentration of a particular RNA is the same in both samples, you'll get a yellow signal.. if one sample is stronger than the other you'll either get a red or a green signal. So the ratio is relative gene expression levels.

 

[Monique]

You can make your own micro arrays, but there are also custom arrays available and even ones with every single gene (about 30000) of the human (or other) genome on it. Besides that they are working on making protein chips, with many different epitopes on them.

 

[karthik3k]

Actually my work is on the analysis of the output.
Now the problem is where can i get the output ?
and in what format ? (intensity/ratio/ ... ) ?
Any links ?

 

[Monique]

Depends on the kind of array you are working on. There are house-keeping genes on chips that allow you to correct for the difference of total RNA of the two samples that went into the procedure originally, the expression level of the house-keeping genes should be the same.

You probably have two samples that have a different biological source, like diseased vs healthy tissue. You'd be interesed in the ratio between the two samples, not absolute values.

If you are looking for a biological marker for disease phenotype, it could be interesting though to specifically look for genes with high intensity and different ratio, since those are easier to detect. If you are looking for the biological foundation of a disease phenotype, it might be better worth looking at low expression genes. Those are more likely to be downstream of the signalling pathways and thus be the cause of other disease markers to be upregulated much more.

What kind of chip and what kind of program are you working with?

 

[karthik3k]

I just want the data file...

Give me some URL to download ...
No matter what kind of chip it is ...

 

[Monique]

Good luck looking for it yourself.

 

[Moonbear]

You're not going to find such a data file on the web. Anyone using the chips for research isn't going to put their raw data out on the web for just anyone to use. You might be able to contact the company that makes the software you're using to get a sample data set to give the software a "test drive", but all the software is copyrighted, and the data files probably are obtained in formats specific for the software bundled with the chip reader.

 

[Monique]

Actually I think journals were planning to make available raw data from experiments published by them through the web (since published data should be free for everyone to look at).. look up a Nature article with microarrays and see if there is extra online content.

 

[karthik3k]

Hey thanks guys.
I found it in Stanford Microarray databse.

Thanx for the those theories...

BTW, Can we work with more than two genes at a time ?? I mean more than 2 dyes .. ??

Coz, tumors may produce many genes...
To identify them or to see the level of expresion of those genes...

 

[Monique]

The two dyes are not genes, but samples. Affymetrix has human genome chips which carry all 30000 some genes, this method uses biotin labeled probes though and not fluorescent dyes.

You could use three dyes, but there would be overlap in spectrum and you will get too much competition of the sample for binding to the probe (be careful, probe and target are often switched around in chip lingo, not sure myself which is which).

 

[karthik3k]

So you mean ...
Its Ideal to use only 2 samples and one gene in MICRO ARRAY ?

or

we compare only relative expression on of single gene in two samples ?


So what does rows and columns in Micro Array represent ?

 

[Monique]

Every spot on the microarray represents a single gene or transcript, they are short uniques pieces of DNA to which another piece can bind.

For instance, you obtain biological samples of a patient and a control, you isolate RNA from that, you make cDNA and you label that with a dye. The patient cDNA you label red and the control you label green. You then hybridize this cDNA (which represents the gene pool that is being expressed in the sample) to the array with all the different pieces of genes spotted on it. You then wash away everything that doesn't bind. If a spot lights up, it means that in the expressed gene pool the particular transcript was present. If it lights up red, it means that transcript was expressed higher in the patient than in the control. If it lights up yellow, the expression levels were the same in both samples.

 

[Monique]

There are all tricks on microarrays too, to determine the aspecifical binding and also the ratio of 3'/5' transcripts, not sure if you want to know..

 

[karthik3k]

hey thanx.
Got a doubt ...
Do you label all the cDNAs from a sample ?

If yes ! , how do you know which one(gene) glows ?

Eg. THERE are N genes (probe) in slide. And u add P cDNAs to the slide. If u label all.
How do we know whether its P1-N1 or P4-N4 ... ?

Im really confused day by day .
PLease help me !

The following Link confused me a lot
http://genome-www5.stanford.edu/cgi-bin/data/grids.pl?fullID=21508GENEPIX21508
its from Stanford Microarray Database.

When we click on the spot on the image it shows the information about that.

Different spots shows different genes.

BTW,
What are 5' and 3' transcripts ?

 

[Monique]

An array is spotted by a robot, you have a platform with a lot of pins, each pin goes into a solution with a different DNA mix, a drop is transferred to the array slide. It is exactly known which spot is which DNA, there is software that places a grid over the spots and can read all the values of all the different spots and since it is an ordered matrix, gene names can be assigned to each spot.

Spot A1 will always be ACE, spot A2 will always be TULP for example.

So yes, you label ALL the cDNA (made from RNA) of a sample, depending whether the ACE gene spot lights up or the TULP gene spot, do you know which was in the sample and how much.

Ofcourse it is possible that the TULP gene spot binds other cDNA than it is supposed to, aspecific binding. Commercial arrays measure this aspecific binding by including for instance 12 different probes for each gene. Some probes have mutations, if something still binds to it, it is aspecific and the signal should be subtracted from the signal of the original probe. But this all happens behind the scenes and you don't have to worry about it, but it is good to know.

With the 5'/3' transcript ratio you measure the extend that your RNA template was degraded and how well the amplification procedure went. I think that if the amplification was not efficient, you will be losing the 5' end of the RNA. This is also important to know, since if the piece of DNA of the probe is in the 5' end, and you lost that during the sample preparation, you won't get a signal, while the 3' end of the molecule is still there. Commercial chips will use probes of the 5' end, 3' end, and the middle of an RNA to get an accurate reading (they are several out of those 12 probes I mentioned earlier).

 

[Monique]

So: the DNA of the array is all synthetic and you know exactly what they represent. You throw on a mixture of two sample labeled with different colors. Depending on which spot lights up in which color, do you know what was in your sample

 

[Monique]

http://www.affymetrix.com/technology/design/index.affx"

Before selecting individual probes from either exemplars or consensus sequences, the 5' to 3' orientation of each transcript must be determined. Affymetrix uses computer algorithms that combine information from public annotations with in-house identification of splice signals, polyadenylation sites, and polyadenylation signals to distinguish sense from antisense strands. If the orientation cannot be determined unequivocally due to contradictory information, then the probes for both strands are generated.

In general, 11 to16 probes are selected among all possible 25-mers to represent each transcript. In addition to choosing the probes based on their predicted hybridization properties, candidate sequences are filtered for specificity. Their potential for cross-hybridizing with similar, but unrelated sequences, is evaluated.

To obtain a complete picture of a gene's activity, some probes are selected from regions shared by multiple splice or polyadenylation variants. In other cases, unique probes that distinguish between variants are favored. Inter-probe distance is also factored into the selection process. Probes are 3'-biased to match the target generation characteristics of our sample amplification method, but they are also widely spaced to sample various regions of each transcript and provide robustness of detection.

It all depends on what you are doing, but generally when you amplify RNA you loose the 5' end of the transcript, that has to do with the way the primers anneal. For the 3' end you can make a primer with a poly-A tail, so you will always amplify the end. For the 5' end there is no common tail for all the different gene-transcripts, so a random primer is used. That primer can anneal anywhere in the transcript so you never know how full length your cDNA is. That is why there is a bias towards the 3', it depends how efficiently the amplification procedure went, whether there are long amplicons.

So for instance if you're working with a custom array that was made and the probes are all from the 5' end of the target, there is a bias that signals can be missed. I'm showing that in order to analyze data, you need to know how they were generated in detail, how was the RNA amplified, how were the probes made, is the information acquired representive of the original material? .. science :)

 

[karthik3k]

What are Predictors ?

How do i find them in Micro array clusters ??

How do i analyse the clusters formed by Micro Array data analysis ?

 

[Monique]

Predictor: a reliable marker that allows you to give a prognosis or diagnosis on a patient for instance. You must know the answer to the other questions.

 

[karthik3k]

Iam sorry. I didnt get it :(
Can you explain in few lines please... ?

 

[Monique]

What exactly are you doing? What kind of experimental data do you have? Disease vs Healthy, Untreated vs Treated, Stage 1 vs Stage 2..?

A predictor would be the gene that is expressed in a different way between the two samples. If it is always red in the disease array and always green in the healthy array, you have a marker for the phenotype.

 

[iansmith]

Predictors could be replaced by the terms negative and positive controls. Negative control is a reation that will yield a negative (no) result and positive control will yield the desired result. Using theses control, you will be able to differenciate between results and interpret the data to reach a conclusion, prognosis or diagnosis.

 

[Monique]

I don't think they can be replaced by the terms negative and positive controls. On an array negative controls are bacterial genes and positive controls are house keeping genes, to make sure that there are no foreign DNA contaminations and that you can standardize the data between arrays.

Predictors are differentially expressed genes that correlate with the phenotype.

 

[iansmith]

Predictors are differentially expressed genes that correlate with the phenotype.

It is still a control. Some people referred to these as background control and it migth be a more suite term. Positive backgroung for reaction that fits the desired phenotype, negative backgorund for the absence of the desired phenotype. Negative and positive control have to be specified for every experiment, it is not set in stone. Negative control is not always to test for bacterial gene and positive is not always for housekeeping gene.

 

[Monique]

How can something that you need to verify experimentally be a control? A background control?? Unless you already have a marker for disease and you want to find more.

Could you give a specific example?

 

[iansmith]

Let say you want to test if an unkown cell share similary with a malignant cell. First you will take cell that you know is healty; this becomes your negative background control. Second you take cells that are knwon to be malignant. This is your positive background control. You also need control to test for contaminations (negative control) and the quality of your experiment (positive control). You run your experiment and compare the sample to the background control and you make conclusion.

 

[Monique]

Ah ok, you are talking about individual samples :) Your set-up might work if you are only testing a handful of genes, but when you are analyzing 30,000 genes that surely is insufficient and won't work.

Also, from karthik3k's question I think we are talking about actually finding genes that act as markers

What are Predictors ?

How do i find them in Micro array clusters ??

Karthik3k, we need you to clarify what you are doing! :)

 

[Moonbear]

I think some of the confusion here is the difference between a methodological control (one that shows the array worked) and an experimental control (one that shows your samples were collected properly and no unexpected variables were introduced). A methodological control would be those Monique is talking about, genes from another species that are or are not added to a sample and should show up present or absent, respectively, on the array. If these don't show up when they are in your sample, or do show up when they aren't included, then you know something went wrong with the array, such as contamination of some sort or one of your fluorescent tags bleached. An experimental control is what Iansmith is talking about, genes that have been independently verified using other methods (real-time PCR, for example) to be different for your two groups (i.e., healthy vs diseased tissue, or pre- vs post-treatment with a drug) in a known direction (i.e., increased following drug treatment compared to pre-treatment). If these don't change in the predicted direction, then you know something is wrong with your sample, such as a mix-up of what drug they were treated with (as one of my former professors was well-known to ask during qualifying exams, "How do you know you didn't just inject flour?"), or the drug went bad, or some unexpected variable was introduced. Every experiment needs both types of controls, methodological and experimental.