Designing effective experiments

As scientists, we are faced with doing experiments almost everyday. So one would presume that we should be experts at designing and performing experiments. But yet every scientist I know has lamented about a failed experiment, wasted reagents and money down the drain; countless hours spent on analysing reams of data, only to have it come to an inconclusive end. I do not claim to be the best at designing effective experiments, but here are some tips I have learnt over the years that might aid you in your attempt to perform an effective one.

1. Having the goal in mind

Having a clear experimental goal is essential. What do you hope to prove beyond a shadow of a doubt? And how do you hope to prove it?

It is important to have some background knowledge before attempting any experiment. Classic sources in addition to your own knowledge are scientific papers (to see what others have done), your supervisor/colleagues, and the world wide web.

A typical goal would be to show that thing A does something in Function B. To do this we can remove thing A. If thing A is a gene, start looking at RNA interference assays or gene knockout techniques. If thing A is a compound, test the compound in different systems to see if it affects function B. If you manage to successfully remove Thing A, try restoring it to see if function B returns, typically known as a “rescue” experiment.

Another way to go about it – Has thing A always been implicated whenever function B is occurring? Look at whether thing A is present whenever function B is happening. Or look at case studies where people having a dysfunctional thing A while having function B compromised. Note that all this is only correlative though and actual experimental-driven evidence can only be obtained by disrupting thing A.

Having clear goals also ensures you do not embark on a pointless experiment. I once saw a poster with the objective to prove sleep deprivation is detrimental to attention. Now I am pretty sure we do not need experiments to prove that!

2. Having the right controls

This is the most neglected aspect of many experiments but often the most key. It tells you when the experiment is working as it should. Positive controls are conditions where you know you will obtain a positive result, negative controls are the converse. Many scientists seem to have the idea that all their experiments are ALWAYS working as they should, which is often not the case. Sometimes the machine has broken, sometimes the enzyme has gone bad, sometimes someone is sabotaging your experiments, you just never know! So a great way to have peace of mind is to have controls, at least the positive one.

3. Number of conditions to test

This is probably one of the things that scientists spend the most time deliberating on and can vary with our personalities. Some scientists like to test a wide array of conditions at a go, some scientists like to test one condition at a time. There are advantages to both schools of thought.

Testing more conditions lets you learn more and could also save you time especially if there is a long waiting time to results. Some things to take note of – equipment capacity and your organizational ability. If you only have a certain number of reaction chambers, there is no point in preparing more tubes than necessary. And if you spend too much time preparing such a vast experiment, oftentimes variability gets you. Letting one tube sit out on ice for a few min longer than the other tubes just leads to greater variation in results. You also run the risk of dropping everything on the way to the machine. So try to keep things simple. Testing one condition at a time is a safe way to do this, though sometimes I get a little too impatient for this. It’s a fine balance. When doing experiments for the FIRST time though, it is ALWAYS best to start small.

4. Repeating the experiment

What is science without statistics? And we all know that for something to be statistically significant, we need an n of 3. So do not hesitate to repeat the experiment to confirm your results. And do not jump to conclusions based on a single experiment. If you are getting a different result each time, you might want to think about those controls!

5. Knowing what to do when your experiment fails

So your experiment is failing. If your positive control is also failing, make sure you go through the protocol carefully. Are you following every step on point? Understanding the protocol (reason behind each step) helps tremendously in troubleshooting and makes you a better scientist. At times it may even spare you from unnecessary cost/steps.  If you have the time, take apart each variable and test what went wrong. If not, people like to try everything with new reagents. Occasionally, this works.

So what happens if your positive control is working but you are not getting the results you were hoping for? Many at this point would keep repeating the experiment, tweaking the conditions a little till they get the result they “want”. This is where most reagent wastage happens. Ask yourself though, are you an open-minded scientist? In other words, are you open to the possibility that your hypothesis might be NULL? NO, no that would never happen. And behold, the data fabrication and paper retractions commence. A true scientist would be open to the possibility of an alternative hypothesis.

So it’s back to the drawing board. Re-think, re-hypothesize, re-plan, re-do. What if, everyone else has proved it but you are unable to? Question the data, question the method, question the reagents, question the experts. Try to prove the hypothesis with a different method.

6. Recording, interpreting and presenting the results

Often key to determining future design of effective experiments. One has to be diligent with recording of data. Have I not said this before? Without proper records, time can be wasted repeating the same experiment or forgetting why the experiment was done in the first place!

Take time to understand the results and discuss them with your mentors or colleagues. People often have different takes on the same piece of data. And often we can be pretty biased about what we want to see, so you might want to try “blinding” yourself i.e. getting someone else to label the tubes so you cannot predict the results.

Presenting the results to an audience, be it a single person or a crowd, also helps to collect your thoughts and formulate future decisions. It is also good practice for your scientific communication skills.

7. Taking it to the next level

Some ways to increase the impact of your experiment:

  • Use a novel/ingenious method
  • Show it occurs in different biological systems
  • Find a real-world application
  • Change the way things are currently done

So go on, what are you waiting for? Design your best experiment ever.