Researchers at University of Maryland have used an artificial intelligence (AI) system to create an abstract language from the constant motion of biological molecules.
Proteins and an AI system
The molecular function of a protein is often determined by its shape and structure. Therefore, exploring the dynamics which control these features can aid our understanding of how proteins work and cause disease, and – importantly – of how to design better targeted drug therapies.
Recurrent neural networks (RNNs) are a machine learning/AI technique for modelling temporal sequences. Long short-term memory (LSTM) neural networks are a popular form of RNNs, used for challenging tasks such as language modelling. Recently, LSTMs have been shown to have the potential to mimic trajectories produced by experiments or simulations. Thereby, making accurate predictions about a short time into the future.
LSTM model
In this study, published in Nature Communications, researchers used LSTM to perform kinetic reconstruction tasks. Specifically, they showed that a simple character-level language model based on a LSTM neural network could learn a probabilistic model of a time series generated from a physical system. They found that not only could this model learn the Boltzmann statistics (distribution of particles over various energy states), but it could also capture a large spectrum of kinetics.
By applying natural language processing tools, researchers were able to create an abstract language. This language described the multiple shapes of a protein, and how and when it transitions between shapes. This is the first time a machine learning algorithm has been applied to biomolecular dynamics. The team believe this represents a stepping stone for using RNNs in understanding the dynamics of complex stochastic molecular systems.
Pratyush Tiwary, senior author, stated:
“Here we show the same AI architectures used to complete sentences when writing emails can be used to uncover a language spoken by the molecules of life.
We show that the movement of these molecules can be mapped into an abstract language, and that AI techniques can be used to generate biologically truthful stories out of the resulting abstract words.”
Image credit: By selvanegra – canva.com
