A research group led by Tomoei Takahashi, a doctoral course student at Nagoya University, succeeded in deriving the world's first mathematical formula for protein design without simulations or experiments.This mathematical formula eliminates the need for simulations that require a huge amount of calculations, and enables overwhelmingly high-speed design.
In the field of drug discovery, new proteins are designed to exhibit necessary functions and efficacy.In the past, experiments and structure search simulations were performed while gradually changing the amino acid sequence, which is the blueprint for a protein, to check whether the resulting protein exhibited the necessary functions and efficacy.Accurate design requires a huge amount of calculation.
Last year, the research group combined a new hypothesis on protein evolution with a machine learning method, and succeeded in realizing protein design overwhelmingly faster than conventional methods.Furthermore, by applying the theory of complex system physics and information statistical mechanics, we succeeded in deriving a mathematical formula for estimating amino acid sequences that does not require simulation.
As a result, the time required for protein design has been reduced to about 10/1 compared to using the efficient design method announced last year.In addition, since the obtained mathematical formula is in a format that allows parallel calculation for each amino acid residue, there is a problem of exploding computational complexity when applying it to the design of a huge protein, which is the target of actual drug discovery. It also has the advantage of being less likely to occur.Moreover, since the theoretical model used this time does not depend on the detailed properties of proteins, it is expected to be applied to problems such as the design of new materials and new devices.
Paper information:[Journal of Statistical Mechanics: Theory and Experiment] The cavity method to protein design problem