Maybe you’ve been optimizing an experiment for a while and finally have it working. You do that first solid trial and see a cool effect! Hurray!! Now what? The urge to run and tell your advisor might be strong, but you can definitely anticipate some of their next questions. Most people will tell you not to get too excited until you repeat it, but let’s go over some specific steps you might take to validate your result.
1) Think back to your experiment and whether the conditions and procedure went according to plan. This information should be in notes you took while performing the experiment. Were your cells healthy to begin with? Was the timing of the experiment precise? Did something strange happen during the procedure? If something went wrong that might affect the outcome, repeat the experiment.
2) Have you included relevant positive and negative controls in your experiment? Did they work? REALLY don’t bother to show it to your PI if you have omitted controls or the controls don’t work. Neither you nor your advisor will be able to interpret the result without working controls. Importantly, come up with a theory for why the controls didn’t work, fix, and repeat. Don’t just do the exact same thing again hoping for a different result unless you have thought about this first or have reason to believe something might have gone wrong (see #1 above).
3) Have you quantified the data? Don’t go with your gut or trust your eyes! Most of the time we are not capable of distinguishing by eye and are often biased by what we expect the result to be if not blinded. If the result is as unambiguous as night and day (there vs gone, many vs none, completely in the wrong place etc), go ahead and do a happy dance. Your PI might still not be as excited as you hoped if you haven’t repeated it, but in my opinion, a very solid trial with working controls and a quantified outcome is worth discussing. If there MIGHT be an effect but it’s hard to say, start quantifying any properties that might be relevant, even if it’s not what you were after initially. Sometimes what you were looking for isn’t true but something else relevant is. Collect some descriptive statistics that might be useful. All this is to say, your experiment isn’t done until you have analyzed the data.
4) Now that you have analyzed the data and know the true result, you’re ready to do a biological replicate (repeat the same experiment with new samples/cells/trials). If you get something different the second time, it’s tempting to do the experiment again to break the tie. A high degree of variation from experiment to experiment might mean that you haven’t controlled for some critical variable or that the parameter you’re testing is not the predominant causative factor for the process of interest. It is worth re-thinking both the design and premise of your experiment. Definitely tell your PI if the result is repeatable. They are likely holding their breath and will be ecstatic.
5) Finally, if the result is repeatable, what does it mean? Your experiment is also not done until you have interpreted the result. Make a model of what you think is happening biologically with your new information. Now test and try to break the model using ORTHOGONAL experiments. This is key. You can do the experiment 1000 times, but it can’t give you the confidence of multiple orthogonal results (particularly if the effect size is small or there is some variation). If the results or orthogonal experiments are consistent with your initial finding, this is the home run. Your result is solid and you’ve found something new! Congratulations!! You and your PI can both do the happy dance! If the orthogonal results aren’t consistent, it’s time to revise the model and come up with a new set of experiments to test it. This isn’t a failure. Herein lies the real beauty of science: systematically getting ever closer to discovery!