1: Protein Evolution

Because this is my first entry, I'd like to just mention: Many times, I will be typing phrases directly from the articles or summarizing viewpoints and conclusions that I don't claim to be my own. Also, I hope to get a lot better at this (in numerous ways) as I practice more and time progresses.

Importance
So, this is a technique that, if people can get it to work well, enables several important things:
1) Re-engineering of natural enzymes and pathways to accept novel reactants or produce novel products. Think about creating new antibiotics, at the very least!
2) Creating completely new proteins with previously unheard of functions, catalyzing unprecedented chemical reactions, eliminating the need for hazardous chemicals and improving the yield of synthesis for many important drugs.
3) Exploring the protein folding problem and linking sequences to structure and function.

Summary
We can harness the principles of evolution (selection for a favorable trait or ability, then replication/proliferation) to engineer biomolecules in directed ways, specifically proteins and nucleic acids. These types of polymers are inherently combinatorial, assembled from their monomeric building blocks, and are thus suited to this type of approach.

The basic strategy is to first construct a library of sequences exploring biochemical space, usually through modification of the gene encoding the protein of interest - this can be done using random or targeted mutagenesis, DNA shuffling, and degenerate codons. After creating the library, you can then screen it or use selection, using a variety of techniques lke HPLC, NMR, Mass spec, detection of color/fluorescence, and FACS.

So, there are 3 cool techniques that help this to be feasible: phage, ribosome, and mRNA display
Phage display: Fuse your gene to a bacteriophage surface protein gene. When the virus particles lyse the host, they will display the protein of interest on their surface, and carry the sequence encoding that protein with their DNA inside.
Ribosome display: Attach a spacer sequence right before the stop codon of your protein, so that it gets stuck inside the ribosome as an mRNA:ribosome:fully-folded protein complex
mRNA display: fuse puromycin to the end of the mRNA for your gene of interest. This will get caught in the ribosome's A site and get fused to the protein. So, you get an mRNA-protein hybrid.

Whichever technique you use, you could then use affinity chromatography to pull out the protein that works (for example, if you're looking for a protein that can bind ATP, attach an ATP-analogue to the resin, and it will pull out any proteins fused to their encoding sequences)

The advantage of the latter two is that you can express entirely in vitro without cells, and still retain the sequence if you find a protein that does the job.

You can either: 1) design from scratch or 2) redesign an existing protein

1) Designing from scratch requires little information (if you want to create just something that folds or has a specific secondary structure), but it's difficult to engineer something that binds a small molecule (20 out of 10^13 in an 88aa library could bind streptavidin, and with little affinity). They've also created ATP-binders, which had adopted a fold fiound in native proteins. They can create alpha-helices/beta-sheets by predefining sequence positions of hydrophilic/hydrophobic residues (some even bound heme!)

Computational screening has enormous potential in allowing virtual screening of many more sequence combinations. The program alternates between sequence design and structure prediction to evolve a molecule over computational time.

2) Redesigning can help improve enzymatic stability, tolerance for different conditions or substrates, and introduce the ability to produce new products. In one cool example, they engineered Cre recombinase to excise out HIV provirus from the human genome

The feelings from many scientists that I've spoken to on this are:
-it's only worked for a few examples. And there are a growing number of supporting experimental successes. But could you do this for enzymes that are not so well behaved? Not yet.
-It's all about your ability to come up with a successful screen. You need to wittle your way down from an incredibly large library to something that has the desired function. According to this article, the biggest challenge is that functional sequences are rare compared to the number of theoretical combinations.
-I also have a problem with the "rational" in rational design. With the whole randomized mutation technique, you're kind of waiting to stumble upon a sequence that works. How good are they at determining specifically, "Ok, if I change this one residue, this will happen..."?

This review really didn't help answer that question. :/ But I have another one waiting to be read, and maybe that'll give me better insight. Sorry, this wasn't a good first entry, hopefully things will get better.