I made an approximation of the microtubule binding domains of kinesin (the red dot is ATP).
I've actually been wanting to crochet proteins for a while, but there is a serious activation energy (ho ho!) involved in trying to replicate a complex three dimensional shape. I ended up just sitting in front of the computer with JMol (the viewer from the Protein Data Bank) open. I made a few cartoony sketches using the handy view presets. Then, as I worked, I tried to ignore all the crazy details that I hadn't incorporated into the sketches.
At this level, the kinesin heads are pretty much slightly curved, tapered objects with lots of bumps on them. One thing that was helpful in deciding how to form those was this little tool I made:
Each nub is made by a set number of increases followed in the next row by the same number of decreases, which are reported in felt. This way, when you are looking at the pdb file, you can easily estimate the size of "bump" you are going to need.
After I was finished crocheting, I used the ends of the yarn to "cinch" the structure together: I tucked the yarn inside the head, pulled it out on the opposite side, pulled it as tight as I needed to so that the structure created the pockets and bends I needed, made a knot, and then tucked the yarn back inside. All three axes were cinched in this way.
Then, after adding some red ATPs to the nucleotide binding pocket, I just sewed the coiled coils together with some like-colored thread. By the way, it's a good thing that so many proteins have coiled coil domains because they are Dead Easy to crochet.
This was really fun, and without question the most challenging thing I've tried to crochet. I'll definitely try more molecules in the future. I hope you do too - let me know how it goes!