Week 13: Biosensors

Summary

Protein science is entering an exciting new phase enabled by major advances in computational and experimental methods. Computational protein modeling has achieved unprecedented accuracy in predicting protein structures at atomic-level accuracy, and designing proteins as biocatalysts, with new binding partners (protein-protein, protein-DNA, protein-ligand) and self-assembling materials.

On the experimental side, with next generation technologies to synthesize and sequence DNA, we can evaluate millions of protein designs in a single tube, and quantify function by high-throughput sequencing. By combining computational modeling and high-throughput experiments, we can design, build and test a large number of candidates, and select the best performing design.

Biosensors are genetically encoded sensors that are specific to the target molecule and report concentration in each cell, allowing us to select the highest producer by high-throughput sorting or genetic selection.

In class, we learned computational methods for modeling protein structures and interactions, including the tools in Rosetta protein modeling suite. It works by creating hundreds of possible folded possibilities and assigning each an energy score. Plodding a slow walk toward the lowest energy form.

A gamified version of Rosetta is FoldIt - thr protein folding GAME! Computers are good at large scale topology computation, but in all honesty, there is indispensable value to human intuition. This game you can manipulate and fold the proteins with your own hands, while you compete with high-scorers around the world to discover the protein structure!

The homework for this week is to run an ab initio folding algorithm in Rosetta software suite and compare the resulting models to native structure.

Computational Assignment

1. Pick a protein to perform ab initio folding.

First, we must pick one of the test cases to run structure prediction calculations. I picked 2HFQ as I found it really intriguing that we still don't know the function of this protein!

Fig.1: I picked the protein 2HFQ.

2. Generate models using AbInitio folding in Rosetta.

Then, we set out to generate models using AbInitio folding in Rosetta. As the code was already provided to us , we just had to execute it:

$ ../executable/AbinitioRelax.static.macosclangrelease @abrelax_flags

The program was running until it generated 100 models, which took a loooong time (~7.5 hours on Core i5, 8GB ram).

Fig.2: Native conformation of the protein in PyMol viewer.
Fig.3: Energy vs RMS plot of 100 generated models.
Fig.4: Comparison of minimum energy structure to native.
Fig.5: Comparison of minimum rms structure to native.