Protein Design Is Inverse Structure Prediction
22 March, 2007 - 2 min read
If we imagine protein folding as the exploration by a human explorer navigating a rugged landscape in search of the lowest elevation point, then how do we describe protein design? If we are still baffled by structure prediction in the ‘midnight zone’ how can we have a bioengineering strategy that is worth the title ‘protein design’?
When does Bioinformatics become Bioengineering? Is protein design really just inverse structure prediction? Does that work to simplify the problem?
I have always looked at the problem from the perspective of predicting ternary structure from primary structure. It almost seems more practical to flip the problem around. Given a structure and function, can we predict the amino acid sequence?
DNA synthesizers are rapidly improving. We still do not understand how a change in the amino acid sequence affects changes in protein structure. Template based modeling is progressing, anything beyond a sequence identity of 30% can be reliably modeled with a good pipeline.
Reliable function prediction is obviously the next milestone, but I think the gap from structure to function is still enormous. How do we standardize function specification? Enzyme functions can be described by the reaction they catalyze or the pathway they are part of. However not all proteins are enzymes. And not all enzymes are proteins. Hormones, transcription factors, membrane receptors all have some characteristic structure that supports their function.
Function prediction needs about 10 years before it can provide the type of infrastructure for robust bioengineering and truly rational drug design.