- Optimization is taking something that works, and making it better.

- Optimization means, to me, finding the optimal conditions to perform your experiment. And these optimal conditions can be the amount of the reagent that you need to use, it can be the number of cells that you need to use in your particular experiment, the number of mice that you need to use. It means trying to minimize also the cost when you are doing experiments, trying also to be efficient. You don't want to find yourself handling 50 mice at the same time. So it's very important to optimize in advance, to estimate which are going to be the conditions that are going to give you an outcome and at the time how many mice and the cost and the time of the experiment. Those pilot experiments that you do to figure out the proper conditions to do your experiments are the optimization process.

- I think students, and scientists more generally, should care about optimizing their experiments because we really want to make sure that the experiments that we do are able to give us clear and convincing answers, when those answers exist. So if you carry out an experiment, and you're expecting a particular effect from your treatment, you want to make sure that the experiment that you do actually detects that effect with a reasonable frequency.

- To me, optimization means that as I design an experiment, there might be a signal-to-noise ratio, a noisiness of this experiment that I can't get at the true measurement that I'm trying to emerge out of that experiment. So maybe I will tweak the experiment by moving it to a different temperature, or to even a different time of day in some daylight cycle experiments, it may matter for that in biology.

- The signal-to-noise ratio compares the level of an experimental signal to the level of what might be considered a background noise in your experimental measurement. So just for example, if I want to measure a signal from a GFP labeled protein but there may be some background auto-fluorescence in that cell. So obviously if that signal is comparable to the background fluorescence of that cell, it's going to be very hard to distinguish and make an accurate measurement.

- So if you have a really noisy assay, let's say, or something where the signal-to-noise ratio is quite low, you might do the experiment one time, and you'll see between your two treatment conditions of interest, you have a very large difference. And you find that really compelling, really interesting, and you want to publish it, right. But is that difference due to actual real biological processes that are differing in these two conditions, or is it just because they are completely randomly varying and you just got lucky this one time. If you just got lucky, that is probably because you have a really noisy assay. So when you repeat the experiment, if it is a quite noisy assay, this difference might reverse, it might shrink. So the influence of the noise is going to be quite high there. So by maximizing the signal-to-noise ratio, you are increasing your confidence that if you get a result once, it's going to be consistent and trustworthy.

- So you'd like to be in a situation where the signal is well above that noise, so you can clearly distinguish what you're interested in. So there are two ways to do that. One is potentially to figure out a type of strategy that decrease the background noise level, and it's worth spending some time in your experimental design to think about whether one can do that. Because a higher signal-to-noise simply means more accurate type of measurement. But in some cases in biology the signal is very close to the noise level and one shouldn't be afraid of tackling that situation too. It's just that one may have to tackle it in a very different way. But no doubt, if one can get a higher signal-to-noise ratio, that makes it much easier to interpret and analyze your data.