Experimental Design – Part 1 – Hypotheses

Note: If you prefer this information in text form, you can reach those pages here

Scroll to the bottom of this page if you want to work through these slides on your own or view them on slideshare via this link

Reach part 2 via this link

Test your knowledge….

miRNA21  has been shown to be upregulated during the contraction phase of wound healing. Which of the following options are appropriate hypotheses to follow up this observation. There is more than one correct answer.
Fibroblasts transfected with a miRNA21 mimic will contract collagen gels more than fibroblasts transfected with non-targeting miRNA mimic controls.
Yes, the observation is a correlation, the increase in miR expression could be a cause or could be an effect or could be unrelated to the contraction. This hypothesis would work; it’s specific and testable.
Is miRNA21 expression decreased in patients with diabetic ulcers
This is a reasonable question but it’s not a hypothesis!  A hypothesis is the answer to the question, a statement that your experiment will test
Treatment of slow-healing wounds with miRNA21 increases their closure rate
Yes, this is a testable statement. However, I’m not convinced that you could justify treating human wounds on the basis of just the observation in the question, indeed I wouldn’t grant a human or animal ethics license based only upon that observation. However, ultimately, this is the sort of destination that you would want to end up in; a clinical trial comparing treated vs control treatment and comparing repair rates.
miRNA21 expression is a master regulator of contractile protein expression
This reads like a hypothesis but is it specific and is it testable? You could argue with me that knocking down or overexpressing miRNA21 and doing a RNAseq or proteomics experiment could test this statement. However, you’d need a tight definition of what “master regulator” means. What data would be sufficient for you to reject the hypothesis? If one gene changed would that be enough? Do you need 2, 3, 10? This is specificity that you are looking for in your hypothesis. Being tight at this stage will help you in planning the rest of the experiment particularly in terms of determining groupings, sample sizes and stats tests to use. Note that these questions are designed to reinforce the learning.
You think that lamin A might be controlled by circadian rhythms. For each of the experimental options below identify if the experiment is correlative or manipulative. Before you click the answer, think about hypothesis the experiment will test. NB the circadian rhythm is controlled by clock proteins such as Bmal and Per
Synchronise cultured epithelial cells and then extract protein extracts at different time points after synchronisation. Determine the relative lamin A expression at each timepoint – CORRELATIVE
Yes, the “manipulation” involved in synchronising the culture is just to provide the control, here you are correlating expression to time after synchronisation. Your design will take into account reasons why you might see correlation even it doesn’t exist (reverse causation, confounding variables) and your stats tests will be based around correlative measures. A hypothesis for this experiment could be “Lamin A expression displays cyclical changes in expression following synchronisation of epithelial cells”
 Synchronise cultured epithelial cells and then extract protein extracts at different time points after synchronisation. Determine the relative lamin A expression at each timepoint – MANIPULATIVE
Nope, try again
Compare lamin A expression in cells-derived from a clock protein knockout mouse compared with those from a wild-type littermate control – MANIPULATIVE
In this option, you have disrupted the circadian rhythm machinery and then are doing a comparison between the groups. As written, this experiment doesn’t directly test the circadian rhythms, it only asks if the clock protein influence lamin A expression in any way therefore the hypothesis could be something like; “Lamin A expression is decreased in clock protein knockout mice”
Compare lamin A expression in cells-derived from a clock protein knockout mouse compared with those from a wild-type littermate control – CORRELATIVE
I can see why you would pick this, however in this case you actually have changed something. Here by comparing to littermates you have controlled for lots of the variables and therefore the interpretation of the results can be tighter differences in lamin A expression here are linked to the change in the clock. Click the other answer for the hypothesis
You have obtained ethical approval to study the gait and biomechanics of a remote tribe with the hypothesis that as the members of this tribe do not wear shoes their gait will be different from those that do. You want to set up a correlative experiment comparing. In the list below, I have identified a list of potential confounding variables. Have a think, then click the links to reveal what I thought of.
People in the remote tribe might have distinct diet than the control group. Diet would influence bone and muscle growth, density, repair, development etc
Differences in behaviour between the tribe and control group. For example, if the tribes people rely on hunting and gathering type behaviour and the comparators have different professions, then adaptations to gait may not be shoes but rather what the people do in terms of how much they walk, run etc, how often, how far etc
Genetics. If the tribe truly is remote then likely they will have a certain degree of genetic isolation raising the possibility genetic or epigenetic differences that influence gait, muscle, bone growth etc
Terrain/landscape if the tribe live in different underfoot conditions from the comparators then there would be potential confounder of the different stresses put upon the tissues due to eg up and down hills vs on the flat
There are many more options than the four above. For each of these variables you should start to think about how you might be able to either reduce their effect or how you might identify their contribution to the output.
For each of the experiments below, identify the model system that would be best to study the question using. Before you hit the answer, think about what benefit you would gain or what insight you would lose by moving to more or less complex systems.
  The C52T mutation in cystatin C decreases the rate at which it is secreted
Cultured cells. Asking this question in an animal model would be possible with the appropriate lines but the added physiological relevance wouldn’t actually help you answer the question. The counterpoint would be that secretion rate depends on other aspects of the local environment. Unless you had a specific reason to say why you thought this was the case then, at very least, you should do the in vitro study first.
The expression of pax6 is upregulated in corneal fibroblasts during corneal epithelial wound repair
3D model,  ex vivo model or animal model? The simplest, cheapest and quickest would be to use cells in culture however here you need a system with fibroblasts and epithelial cells that can effective recapitulate the wound healing response. This is definitely do-able but the question you would ask: it accurate enough” On those basis you could justify using a slightly more involved system. Ex vivo systems are taking something that has been alive and maintaining it outside the body. They can be a good compromise; an animal or human may still provide the sample but it could be spare tissue from an animal that would be euthanised anyway. Setting up your study this way may reduce the animal numbers or reduce the suffering of those that you do have to use. If this was good enough to ask your question, then you are ethically obligated to go down this route. However, wound repair involves immune cell activity in response to epithelial damage. These cells release cytokines which fibroblasts respond to, therefore you may need to use a living animal model to answer this question. The teaching point here, is that these questions are part of the design process and you should be prepared to justify decisions that you have made.
  Poly-L-lysine contact lenses have anti-microbial activity comparable to topical antibiotics
in vitro or ex vivo. At least at first, you want a tightly controlled model system. Here you could test the contact lens efficacy directly on cultured bacteria or take something like an eye from an abattoir and infect it in a lab environment. There is no need, at least at the initial stage, to use an animal. To be clear, before you could go into the clinic, your studies would have to progress to animal and then human trials but the specific hypothesis as written above would be done quicker, cheaper, with fewer animals harmed and with tighter more easily interpreted data in vitro.
End-stage renal disease is associated with increased urine creatine
Probably human. Two reasons for this answer; firstly, you can collect urine non-invasively from humans so, although you would still need ethical approval, the experiments themselves would have no detrimental effect on the patients. So, why not use the system you care most about (assuming that is human!)? Secondly, if you were to use an experimental animal system you would be inducing the end-stage disease. This means that the animal would have to be sick for extended period of time. It’s hard to justify putting an animal through this level of suffering, so while it would be possible to test your hypothesis this way, and, indeed, it may give you tighter data, the question about whether you should is relevant to the decision you make.

Reach part 2 via this link