Last week a friend posting this article about failure in science. The timing was perfect because last week was also when I discovered that my project is dead.

It turns out that one of the key tools for my project, the antibody against my protein, doesn’t work for me.

Quick refresher on antibodies: They are natural components of the immune system, huge proteins that have a region that can bind very specifically to molecules (such as proteins). Your body has somewhere between millions and billions of antibodies, each slightly different, and each able to bind to different things. In research we take advantage of this to purify antibodies that very specifically bind to just the protein we’re interested in. This is done by injecting large amounts of your protein of interest into an animal such as a rabbit, mouse, or goat, then letting their immune systems make tons of antibodies against that, then drawing their blood and purifying the antibodies.

Ideally you can inject whole, purified protein (you can oftentimes have E. coli make a bunch of your protein using molecular biology techniques). But because my protein of interest is imbedded in the membrane, it’s much harder to get E. coli to make it, and it’s much harder to purify. So instead the antibody was raised against a small bit of the sequence. So the idea is that the antibody will bind to the ~12 amino acids* of my protein of interest.

*If a protein is a balled up necklace, the amino acids are the beads; they are the protein building blocks.

But my antibody seemed to bind to many other things. So I did some googling (okay, slightly more sophisticated than googling, but I used online search tools to search for that 12 amino acid sequence in all the proteins in Xenopus leavis (the frogs we use). It turns out I got a few hits, so it was likely that my antibody does bind to things other than my protein of interest. Which basically would make all my data messier and possibly unusable.

But the real nail in the coffin came when I decided to look at where exactly the antibody was binding within the sequence in my protein. And I discovered that the answer is…. nowhere.

The sequence used to make the antibody came from Xenopus tropicalis, a close relative of X. laevis. And while their protein sequences are very similar, they’re not identical. It turns out the regions chosen to make the antibody against was almost completely different. So now I have an antibody that binds many things, but does not bind the protein I actually need it to bind.

And there isn’t really anyone to blame. The X. laevis genome wasn’t even sequenced until 2010, and no versions of it were widely available until 2012, and my understanding is that even the most current version has some issues**. So it’s pretty common to use X. tropicalis sequences as a starting point for X. laevis work. This is because tropicalis has a much easier to work with genome. Tropicalis is diploid, meaning it has two copies of each chromosome (one from mom, one from dad; the same as humans). But laevis are tetraploid, meaning they have four copies of each chromosome. This basically just makes the sequence a little messier and harder to identify any give base pair as “correct”, a normal variation (such as a different allele), or an abnormal variation (such as a mutation).

**it’s actually entirely possible that the sequence I’ve used to determine the antibody doesn’t bind is actually itself incorrect, but the fact that the antibody binds to so many other things makes it pretty useless to me anyway.

The moral of the story is that I have a new project! Since I’m late enough with this post as is and because I’m still in early planning and thinking and reading stages anyway, I’ll wait to tell you details until next week. But the basic project is retina regeneration! I’ll be using human induced-pluripotent stem cells that have been differentiated into retinal cell types, and I’ll be studying what helps or hurts their growth and what influences their axon pathfinding.

Also I’ll talk briefly about about human induced-pluripotent stem cells (a.k.a. iPSC). These are cells taken from adults (usually skin cells, such as from a cheek swab) that are essentially deprogrammed through environmental and sometimes genetic manipulations in order to turn them into a stem cell, a cell capable of becoming almost any cell. These cells can’t make a whole organism. They’re pluripotent (capable of making many different cell types) but they are not totipotent (capable of making all cell types). While they’re not the same as looking at actual development (which is where model organisms like xenopus and mouse really come in handy), human iPSCs are a great way to see if the things we discover in animal models hold true for human cells as well without, you know, actually experimenting on people. And the project I’ll be working on is within the context of stem cell-generated replacement retinas that could be implanted in patients with damage or disease to hopefully restore functional sight. So in this context, we actually want to understand how human iPSCs behave themselves (though the things we learn will definitely be more broadly applicable to development, regeneration, and human cell biology).

So I have some pretty exciting stuff to look forward to! Failure’s not all bad. And having heard some other stories, I’m pretty happy it only took 6 months, rather than several years, to figure out my previous project wasn’t going to work. And maybe in a few years when I have a firm foundation with this project, I can look into some new tools and pick up the other one again.