An engineer at the University of Missouri has obtained funding from the US National Science Foundation to develop a revolutionary tool that predicts the function of proteins based on their amino acid sequence.
This breakthrough promises applications from the development of drought-resistant crops to the design of advanced medicines.
From sequence to protein structure and function
Jianlin Cheng, Professor of Electrical and Computer Engineering, presented one of his latest creations, open source software that allows users to input an amino acid sequence to predict both the three-dimensional structure of the protein and its specific function in a cell. In addition, the system is capable of identifying the precise site of the protein where this function is carried out.
Proteins are fundamental to life and their understanding opens doors to numerous possibilities. For example, if a protein that promotes tumor growth is identified in cancer patients, scientists could design drugs that inhibit its activity at the specific site, slowing or stopping tumor growth.
The Power of the Deep Transformer Model
Cheng is using a deep transformer model, similar to the one that powers ChatGPT, to develop his protein prediction tool. Amino acid sequencing is considered the language of biological systems, and the team is creating three types of deep transformer models: one for one-dimensional sequences, one for 2D analysis of protein interactions, and one for three-dimensional structures that consider specific sites in the amino acid sequence. protein.
This initiative represents another milestone in Cheng’s successful career in protein prediction. In 2012, he and his students first demonstrated the superiority of deep learning in predicting protein structures. In the 2020 CASP14 experiment, Deep Mind introduced AlphaFold2, an advanced deep learning method that achieved unprecedented accuracy in predicting protein structures. In CASP15 of 2022, the Cheng-led Research Group further surpassed the accuracy of the AlphaFold2-based prediction.
Cheng highlights the incorporation of advances in protein structure and the use of AlphaFold2 in this particular project. The language model methodology is novel in this field and represents an exciting area of ​​research in which Cheng and his team are devoting much effort.
