计算机科学
计算生物学
蛋白质-蛋白质相互作用
蛋白质组
人工智能
机器学习
功能(生物学)
计算模型
结构生物信息学
蛋白质结构
生物信息学
生物
遗传学
生物化学
作者
Jesse Durham,Jing Zhang,Ian R. Humphreys,Jimin Pei,Nick V. Grishin
标识
DOI:10.1016/j.tibs.2023.03.003
摘要
Deciphering coevolutionary signals in protein sequences and applying deep learning methods such as AlphaFold have led to breakthroughs in modeling protein structures and interactions. The accuracy of interaction partner detection and structural modeling or protein complexes by computational methods now approaches experimental methods, and we are entering a new era where computation will play an essential role in both tasks. We expect rapid progress in characterizing human PPIs, thus enabling biomedical applications such as interpreting pathogenic variants, developing drugs to target PPIs, and designing protein binders to regulate protein function. We still face challenges in modeling transient and weak interactions, understanding the interactions mediated by intrinsically disordered regions (IDRs), expanding to other molecules such as polysaccharides and lipids, and moving towards modeling the entire cell. Protein–protein interactions (PPIs) drive biological processes, and disruption of PPIs can cause disease. With recent breakthroughs in structure prediction and a deluge of genomic sequence data, computational methods to predict PPIs and model spatial structures of protein complexes are now approaching the accuracy of experimental approaches for permanent interactions and show promise for elucidating transient interactions. As we describe here, the key to this success is rich evolutionary information deciphered from thousands of homologous sequences that coevolve in interacting partners. This covariation signal, revealed by sophisticated statistical and machine learning (ML) algorithms, predicts physiological interactions. Accurate artificial intelligence (AI)-based modeling of protein structures promises to provide accurate 3D models of PPIs at a proteome-wide scale. Protein–protein interactions (PPIs) drive biological processes, and disruption of PPIs can cause disease. With recent breakthroughs in structure prediction and a deluge of genomic sequence data, computational methods to predict PPIs and model spatial structures of protein complexes are now approaching the accuracy of experimental approaches for permanent interactions and show promise for elucidating transient interactions. As we describe here, the key to this success is rich evolutionary information deciphered from thousands of homologous sequences that coevolve in interacting partners. This covariation signal, revealed by sophisticated statistical and machine learning (ML) algorithms, predicts physiological interactions. Accurate artificial intelligence (AI)-based modeling of protein structures promises to provide accurate 3D models of PPIs at a proteome-wide scale. the phenomenon where two different positions (residues) in a protein or two proteins reciprocally affect each other's evolution, which usually results from direct contact between residues in the 3D structures of proteins. a class of neural networks that are frequently used in image processing. Neural networks are computational methods inspired by biological neural networks. a branch of machine learning that comprises multiple layers of neural networks. methods that attempt to find a mutual orientation of the 3D structures of two interacting proteins that minimize an energy function over the protein–protein interaction (PPI) interface. energy functions derived from statistical analysis of observed states in existing systems (e.g., experimentally determined protein structures). They are designed to be efficient in computing, and more frequently observed states are evaluated more favorably. the total energy of a particular system computed as a function of the state of the system. proteins that function together, not necessarily through physical interaction. regions in a protein that do not adopt a fixed or ordered 3D structure. a concatenated multiple sequence alignment (MSA) of proteins A and B used as inputs for coevolution analysis or deep learning networks such as AlphaFold. In this concatenated MSA, the homologs of protein A and protein B are paired by being placed in the same row of the MSA. The generation of a paired MSA converts the problem of modeling two proteins to a problem similar to that of modeling one protein. proteins interact through direct binding to each other.
科研通智能强力驱动
Strongly Powered by AbleSci AI