Hi Rakesh,

The experimental duplicate can give you more identifications (when you combine both LC-MS/MS runs) and tell you if your machines work fine. Beyond that you don't need technical replicates. They are not good for comparative/quantitative proteomics.

There is a general rule about replicate choice. If one is able to perform biological replicates then one should use them and not do technical replicates. One gets very large statistical benefits from going from 3 to 6 biological replicates. Beyond that the gains in power are probably not  worth the additional analytical time.

Now if only the biological material is the limiting step and biol replicates are not possible one can/should do technical replicates. Yet again, they would not give you additional statistical power. You can conclude the same things when comparing 2 biological conbditions and 2 biol conditions x 3 technical replicates each.

In this line of thought your statement is incorrect. You don't have 6 of A and 6 of B, but 3 of A and 3 of B.

In summary: Don't do technical replicates (maybe just one to confirm that the LC-MS/MS works fine), do more biological replicates!

Hope that helps.

Ilian

Personally, I would not pool your samples prior to LC-MS/MS. If you do so, your 3 biological replicates are diluted into 1 in reality. This may well have the affect of highlighting the major differences between A and B, but you lose power as N = 1 as far as the stats are concerned.

I don't agree you shouldn't run technical replicates. We need be sure we are talking about the same thing. I do not view your duplicates as technical replicates. To me a technical replicate is when you run the SAME sample twice under identical conditions. If you have the time and you don't have lots of samples they can be worthwhile and certainly are not a bad thing. I do agree you should be using biological replicates over duplicates or technical replicates.

I think it depends on the ammount of sample you have to run LC/MS/MS. Technical replicates (2-3 injections of the same sample) are very important in my opinion to have a idea about your system performance.A pool sample is a very good standart to test your statistics and your system stability since this sample have to be in averidge for all detected parameters. The biological replicates will provide statistical power to proove your markers and are very important for that in addition. To avoid additiona problems I would also recomend to randomize the order of the injections.

The general idea of sample replication is to visualize the variance in your approach. Biological replicates address the wet lab variance, while technical replicates address the variance of your LC-MS method. Of course, injections of biological replicates are also influenced by the LC-MS.

By pooling biological replicates, you will hide and lose the wet lab variance, which is probably not the original idea. For the statstical analysis I would split every sample into 3-5 technical replicates. And yes, you will end up with a lot of injections but improve the statistical significance of your experiment.

The answer to this question really depends several factors. Do you just want to compare detected/not detected between conditions A and B, or do you want to quantify relative protein expression? Is your sample limited, or do you have a lot? Do you have access to the mass spec for a long time, or are you limited by access and/or expense?

If you are just trying to maximize protein ID's, I would pool the samples (one pool for each condition) and fractionate them. Assuming you are confident in your mass spec skills, one or maybe two LC-MS/MS replicates for each fraction will suffice. The second LC-MS/MS injection usually only adds ~15% more ID's, and the third injection adds even fewer, so there are diminishing returns. You get a much better return on your time and money by ensuring that you have lots of sample injected, and that you have decreased sample complexity and/or increased LC gradient time to match the duty cycle of your MS.

If the goal is quantification, replicates (biological, sample technical, and instrument technical) are necessary in order to assess the variance and provide statistical power. The question of biological vs. instrumental replicates is an area of active research, but the consensus seems to be moving toward a preference for biological replicates over LC-MS replicates. The reason for this is that the variance in biological replicates tends to be significantly greater than variance in the MS. We have observed this in our lab, and you can test it yourself: do triplicate LC-MS/MS analyses of three different biological replicates and measure the variance in measured protein quantity for single sample injected three times versus the variance between the three biological replicates. Also, since the goal of an experiment is usually to determine if there is a real difference between two different biological conditions, your priority should be to assess biological variance, not instrumental variance. So assuming you have limited mass spec time, I would not pool your samples.

For some good research on dealing with variation in quantification using label-free LC-MS, see Olga Vitek's research (linked below). Also check out references 14-19 in the paper linked below,

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3489535/

In the posts above a very clear distinction is made between technical and biological replicates. One practical aspect of such analyses is that (depending on the level of replication):

1) variance in technical replicates is based technical variability/noise alone

2) variance in biological replicates is based on biological AND technical variability.

I would not pool the 6 sample of A with the 6 sample of B . if you pool the samples , you can not assess the variance between these two conditions and make your sample more complex for LC-MS/MS(may lost some proteins). if you do not have an access to LC/MS/MS, it is worthwhile spending money for two samples.