Protein Structure Prediction: Challenges, Advances, and the Shift of Research Paradigms

Abstract

Protein structure prediction is an interdisciplinary research topic that has attracted researchers from multiple fields, including biochemistry, medicine, physics, mathematics, and computer science. These researchers adopt various research paradigms to attack the same structure prediction problem: biochemists and physicists attempt to reveal the principles governing protein folding; mathematicians, especially statisticians, usually start from assuming a probability distribution of protein structures given a target sequence and then find the most likely structure, while computer scientists formulate protein structure prediction as an optimization problem — finding the structural conformation with the lowest energy or minimizing the difference between predicted structure and native structure. These research paradigms fall into the two statistical modeling cultures proposed by Leo Breiman, namely, data modeling and algorithmic modeling. Recently, we have also witnessed the great success of deep learning in protein structure prediction. In this review, we present a survey of the efforts for protein structure prediction. We compare the research paradigms adopted by researchers from different fields, with an emphasis on the shift of research paradigms in the era of deep learning. In short, the algorithmic modeling techniques, especially deep neural networks, have considerably improved the accuracy of protein structure prediction; however, theories interpreting the neural networks and knowledge on protein folding are still highly desired.

Analysis of a novel insulin lisargine at the molecular level using X-ray crystallography and biophysical techniques

“…Second, measuring various geometric parameters of the crystal diffraction to determine the crystal system, cell parameters, and the number of independent parameters within the unit cell. Finally, measuring the intensity of diffraction spots and employing Fourier transformation to reconstruct the molecular structure based on these intensities 9 . Since each part of the molecule contributes to diffraction, it is essential to measure the intensity of all diffraction spots to accurately reconstruct the molecular structure.…”

Recent Advances and Challenges in Protein Structure Prediction

“…Reference provides a comprehensive development history of each category of structure prediction algorithms. Dongbo Bu et al summarized the research paradigms to the protein structure prediction problem from different fields, such as the perspectives of biologists, physicists, and computer scientists, etc . With the rapid development and successful application of deep learning in the field of protein structure prediction, current methods tend to leverage the advantages of various methods to solve practical prediction problems, striving for higher prediction accuracy across all types of proteins (FM or TBM).…”

“…Dongbo Bu et al summarized the research paradigms to the protein structure prediction problem from different fields, such as the perspectives of biologists, physicists, and computer scientists, etc. 34 With the rapid development and successful application of deep learning in the field of protein structure prediction, current methods tend to leverage the advantages of various methods to solve practical prediction problems, striving for higher prediction accuracy across all types of proteins (FM or TBM). Consequently, in the past two years, protein structure prediction methods have been broadly classified into two categories: contact/distance-assisted geometric optimization modeling and end-to-end structure prediction using deep learning.…”

“…One category is the end-to-end methods using MSA information, such as AlphaFold2, RoseTTAFold, etc. Another category is the MSA-free end-to-end methods, which usually utilize a pretrained deep learning model trained on large-scale protein sequence data sets to learn the latent rules of protein sequences such as ESMFold and thus replace the coevolutionary information derived from MSA . AlphaFold2 and RoseTTAFold are already widely reported and known in the field and will now focus on the recent emergence of other end-to-end methods and their important conclusions and applications derived from corresponding studies.…”

Sampling Conformational Ensembles of Highly Dynamic Proteins via Generative Deep Learning

“…Experimental techniques determine protein structures and probe their dynamics, and bioinformatics tools such as AlphaFold2 and RoseTTAFold can accurately predict structures from given protein sequences [4][5][6] . However, these techniques typically provide a handful of structures with limited information on dynamics.…”