This week me and my summer student got some interesting data. Her first thoughts were ‘oh no, what have I done wrong’- when I asked her about this I found that in her undergrad most of the experiments they do are designed to give perfect data all the time. If your data comes out bad- you have done something wrong. Science research however, does not work like this.
There is no ‘good data’ or ‘bad data’. You are on the cutting edge of research, and you don’t know what to expect- so if results come out doing what you think then cool, but when they don’t – that is also interesting. And if the experiment doesn’t work…
This got me thinking about some of my experiments: the good, the bad and the ugly…
The “perfect” experiment. The experiment that everyone aspires to. You do the experiment- nothing goes wrong, you get good data, small error bars for all the data, and the data does what you think it should. This has not happened to me so far. There has always been something. I don’t think the perfect experiment exists.
In the beginning two years of my project I needed to express two genes in tomato plants. Before I really got into this though, I wanted to check that co-expression of these genes actually produced what I thought it would.
We have this rather nifty experiment protocol called: the transient expression assay. I am only going to explain the basics- but I will put some references at the bottom if you want to find out more!
This uses the plant virus: Agrobacterium tumefaciens. This virus is super cool- it infects plants, and actually inserts part of its own DNA into the of the plant. We are able to use this to our advantage: we put our genes of interest into agrobacterium, and then we infect the plant (normally tobacco) with it- our genes will be inserted into the plant DNA and expressed.
Here is the nifty part: you get a culture of agrobacterium with the genes you want and put it into a syringe, then you hold the syringe up to the leaf (we use tobacco leaves) and push- and you can see the solution spreading into the leaf. You wait 5 days, harvest the leaves- and via some fancy analysis methods you can find if the product you want is in the leaf or not.
In my case I wanted to express 2 genes (named Stilbene synthase and resveratrol-o-methyltransferase), which are not expressed normally expressed in tobacco. Expression of these two genes would give me two products: Resveratrol, and Pterostilbene. If only the stilbene synthase worked I would just get resveratrol. If both genes worked I would get resveratrol and pterostilbene. If the stilbene synthase didn’t work I wouldn’t get anything. So its a lovely and easy experiment, quick to get the results, and the data is nice and clear. And for me, my experiment ran beautifully. Just look at that peak!
However, these beautiful experiments don’t happen all the time… so, moving onto the bad ones.
There are a few types of bad experiment.
- The one where the equipment breaks.
- The one where it just doesn’t work- for no real reason.
- The one where you made a mistake.
- PCR repeats. (everyone who has done PCR knows this.)
I am going to tell you about my latest bad experiment. The one where I did my calculations wrong.
This actually follows on from my post ‘the anxiety fights back’ its about that scratch assay. The one I did with my summer student.
First problem= the stupid microscope camera couldn’t zoom out enough to take a photo of the whole scratch at T0. So annoying.
Second problem= I told my summer student to add the wrong amount of a reagent. My bad.
Third problem= I made a mistake in my calculations so I over diluted a reagent meaning I added wayyy too little of a reagent.
Luckily there were some things that may save this experiment: the fact that it was a testing new things, so didn’t really follow on from anything, meaning a repeat was needed anyway. It wasn’t the end of the world. Things still happened- which were interesting. I now know how to do a scratch assay and have optimised the protocol. I have now learnt how to do my calculations better, and properly check it all (exhibit A).
These are the really bad experiments. The ones that will haunt you. The ones that really did screw up the overall timeline of everything.
For me, my horror story comes from a tomato transformation. Back in the first year of my PhD I started to do a tomato transformation to get the genes described above into a variety of tomato called Microtom. This was a massive fail-oh my goodness.
Again, I don’t want to give you all the details, but I will describe the basics. For plant transformation you use Agrobacterium tumifaciens as above, containing the genes you are interested in. But instead of injected into a plant, you soak plant cotyledons pieces (plant seedlings when they first start growing have a stem and 2 initial leaves coming out- these are the cotyledons), in a solution of agrobacterium. Over the next few weeks you tend to your pieces of plants, and they grow into a callus- a unspecified growth of plant cells. Eventually plantlets start to sprout, and you can take these and grown them up, if they root they can be planted in soil. If they don’t root they weren’t proper plantlets.
This whole process however, depends on one thing: Antibiotic selection. Alongside the genes you add you also add a antibiotic resistance marker. This means that if the plant has successfully integrated the genes into its DNA it will also have an antibiotic resistance gene. By growing everything in media containing this antibiotic, you will know that anything growing is ‘positive’ it contains the genes you want.
I got right through to the end stage. I had lovely plants growing in the greenhouse- and I just needed to test them and make sure they had all the genes. I tested them, and they didn’t have one gene. Not one. Not even a bit. WHAT.
Fast forward a lot of analysis, general debating in the lab, general cursing, and trying a whole bunch of other primers, we found a clue. The plants were expressing kanamycin resistance. They had the resistance to the antibiotic- just no gene. Why? One of the guys in the lab had a hunch and hurried away to do some extra tests, he came back and told me the news. The microtom seeds we had which we thought were wild type (i.e normal with no resistance) weren’t wild type. They had been labelled wrong and had actually been made from a previous experiment and labelled as wild type. I had wasted 3 months!! I had performed a transformation experiment on tomato seeds which already had kanamycin resistance, and so would grow if they had genes or not!!
But even this awful experiment had a few good things… I saved others in the lab also doing the same ill-fated experiment with contaminated seeds. I now knew how to do all the stages of the experiment, and would easily be able to do it again- this time easier! In the meantime others in the lab had further optimised the experiment- so I could potentially use a better protocol.
The moral of this is that experiments work, they don’t work. They give good results and bad. Even a absolute failure of an experiment normally can help in some way- whether you realise something you did super wrong, or realise the equipment was going a bit dodgy. Science doesn’t run smoothly. There will always be data which makes no sense at all! But, I think at least, its what makes it interesting!
(please excuse the Wiki links- it is really hard to find good simple explanations about these experiments which aren’t papers that you have to be subscribed to..)
If you want loads of science-y detail and can access it: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC150518/
Stable transformation (again you will need access to online journals) doi: 10.1007/978-1-59745-427-8_7.