Proteins are basic life components and scientists are trying to improve or create new essential nutrients which can develop new functions and processes.
Researchers from the United States and Taiwan have produced new proteins with the use of artificial intelligence and have published the results of their study in the journal in APL Bioengineering.
New proteins are designed by mimicking existing proteins or by manually editing the essential amino acids that make them up. This process is very long, time-consuming and based on trial and error. In order to do it faster, researchers have used an artificial intelligence technique called machine learning, which allows parts of the scientific method to be automated by using mathematical procedures.
There are around 100,000 kinds of protein in a vertebrate animal. Their functions are not only determined by their structure, but also by their shape.
“The amino acid string of each kind is folded upon itself in a precise manner, coiled about like twine and crumpled together like a piece of wadded paper,” explains the entomologist Edward O. Wilson in his book Consilience.
“The total molecule bears resemblance to forms as variable as clouds in the sky. Looking at these forms, we readily imagine lumpy spheres, doughnuts, dumbbells, rams' heads, angels with wings spread, and corkscrews.”
The most striking element of this new way to create proteins through machine learning is that protein structures are translated into musical scores. An algorithm has been designed to materialise music from sound waves to matter.
Each of the 20 amino acids that make up proteins has a unique vibrational frequency and each protein’s chemical structure can be mapped with audible representations. Concepts known from music theory such as melody, chord or rhythm were used and the sounds which were generated could be utilised to train deep learning neural networks.
As Markus J. Buehler from the Massachusetts Institute of Technology explains “once the computer has been given a seed of a sequence, it can extrapolate and design entirely new proteins by improvising from this initial idea, while considering various levels of musical variations -- controlled through a temperature parameter -- during the generation.”
Researchers demonstrated that they can design proteins whose nature is yet to be conceived, paving the way for the creation of entirely new biomaterials.
By adjusting the temperature, they can increase the number of variations created by the algorithm, while new mutations can be measured to determine which are more effective as enzymes. The new discovery could also help create new composition techniques in classical music by translating the rhythms and tones found in proteins.
"In the evolution of proteins over thousands of years, nature also gives us new ideas for how sounds can be combined and merged," stressed Buehler.
Reference: "Sonification based de novo protein design using artificial intelligence, structure prediction and analysis using molecular modelling," APL Bioengineering, DOI: 10.1063 / 1.5133026.