Integrated modeling of protein and RNA

The correct functioning of organisms heavily relies on proteins and RNAs, which hold significant roles in governing biological pathways and mechanisms from a molecular perspective, as well as controlling cell function and transmitting genetic information [1]. The research focused on RNA and proteins is currently experiencing a surge in popularity. Significant strides have been made to comprehend the intricacies of protein and RNA structure and function through computational and experimental techniques [2–6].

Researchers aim to identify the associations of functions or diseases from the proteins and nucleic acids. Once the structure is accessible, analyzing the three-dimensional information for annotating functions more accurately becomes possible. Through the integration of experiments, this approach provides an invaluable understanding of biological mechanisms and propels the progress of therapeutic strategies for various diseases. Recent technological advancements have led to an improved version of virus structures and facilitated the development of vaccines, particularly in the ongoing COVID-19 pandemic [7]. Besides, targeting human protein and RNA through drug treatments presents a promising solution for curing cancer. Despite their potential benefits, concerns remain regarding the risk of cardiac toxicity and associated high recurrence rates. This has prompted the need for innovative techniques to gain a deeper understanding of protein and RNA structures and functions.

This special issue centers around innovative experimental and computational biological, chemical and medical research methods. Researchers from multidisciplinary backgrounds share their expertise in modeling proteins and RNAs. They utilize distinct strategies to gain novel functional insights into critical biological systems and help solve unanswered challenging questions relevant to health and disease. This topic encompasses diverse groundbreaking experimental techniques, integrating theoretical simulations and cutting-edge computational. Collectively, these elements contribute to advancing the modeling of RNA and proteins.

It is still difficult to precisely determine the structure and dynamics of complex systems through experimental methods. For example, accurately constructing atomic structures of complexes from cryo-EM maps remains challenging [8]. In the article ‘DEMO-EM2: assembling protein complex structures from cryo-EM maps through intertwined chain and domain fitting’, Zhang et al. [9] introduce an automated method for constructing protein complex models from cryo-EM maps. This approach involves an iterative assembly procedure that uses fast quasi-Newton optimization and Differential Evolution algorithms to combine chain- and domain-level matching and fitting for predicted chain models. Their results highlight DEMO-EM2 as an efficient method to provide a reliable cryo-EM complex structure. In the article ‘Enhancing protein dynamics analysis with hydrophilic polyethylene glycol cross-linkers’, Sun et al. [10] utilized XL-MS experimental techniques to analyze structural complexes’ dynamics accurately. They compared two cross-linkers with different backbone properties: the hydrophilic BS(PEG)2 and hydrophobic DSS/BS3. They assessed their ability to capture protein structure and dynamics through in vitro, in silico and in vivo experiments. Their findings from in vitro and in vivo cross-linking experiments strongly support the enhanced capability of BS(PEG)2 in capturing the dynamic structural nuances of multi-domain proteins. All-atom molecular dynamic (MD) simulations and quantum chemical calculation further highlight BS(PEG)2 intrinsic attributes, including heightened hydrophilicity and increased polarity, which facilitate its proximity to the protein surface. These two works have vastly improved experimental techniques for determining complex structures and dynamic changes. Their valuable insights and contributions have propelled experimental determinations to new heights for applications.

MD technology heavily relies on precise energy potentials to analyze the physical movements of atoms and molecules. When conducting MD simulations, transitioning between states separated by considerable energy barriers can be difficult. In the article ‘Differentiable rotamer sampling with molecular force fields’, Sha et al. [11] have developed a mathematical foundation to overcome this limitation using Boltzmann generators. To address the biases in angle generation and practical deficiencies of Boltzmann generators, they have introduced physics-based geometric methods that allow for unbiased and direct sampling of the rotameric conformations of proteins and RNAs. They demonstrate that the Boltzmann generator approach helps generate biologically relevant conformations faster and more accurately than traditional MD methods. Measuring long-range electrostatic potential poses a challenge when extending MD calculations to larger time scales. In the article, ‘A novel approach to study multi-domain motions in JAK1’s activation mechanism based on energy landscape’, Sun et al. [12] introduced a novel electrostatic potential calculation tool, Delphi. They successfully applied this method to investigate multi-domain motions in the activation mechanism of JAK1. This study presents a promising approach to studying long-range electrostatic potential, a typical example of resolving energy barriers. In the article ‘Computational insights into the cross-talk between medin and Aβ: implications for age-related vascular risk factors in Alzheimer’s disease’, Huang et al. [13] employed atomistic discrete molecular dynamics (DMD) simulations to study protein folding and aggregation. DMD is a highly efficient method that simplifies the energy landscape by coarse-graining it instead of integrating continuous energetic potentials [14]. This approach significantly saves time and computational resources, making it an ideal solution for energy calculations of large-scale and long-term MD simulations. The authors have systematically investigated the self-association, co-aggregation and cross-seeding phenomenon between medin and amyloid-β protein in Alzheimer’s disease, utilizing the benefits of DMD to gain valuable insights into these complex mechanisms. These three works have broadened the usability of MD simulation in the analysis of complex cellular processes and mechanisms, providing innovative, fast and accurate solutions to overcome MD limitations.

The special issue also focuses on developing cutting-edge AI computational methodologies for complex structure prediction. One of the areas that require improvement is the accurate binding site prediction. In the article ‘Deciphering principles of nucleosome interactions and impact of cancer-associated mutations from comprehensive interaction network analysis’, Xu et al. [15] proposed network methods to analyze the experimental binding targets. They conducted a large-scale systematic study of the histone interactions network. Their works have identified the preferred binding hotspots on nucleosomal/linker DNA and histone octamer and revealed diverse binding modes between nucleosome and binding partners. The predicted recurrent cancer mutations have significant disruptive effects on histone/nucleosome interactions and may have driver status in the development of cancers. In the article ‘RNet: a network strategy to predict RNA binding preferences’, Liu et al. [16] developed a machine-learning-based network approach, RNet, to predict RNA binding sites and dynamical binding behavior. The binding site identification algorithm integrates local and global network properties and demonstrates remarkable precision and robustness regarding perturbations. Additionally, the highly effective distance-based dynamical graph algorithm considers geometric network properties to characterize RNA complex interface binding behavior accurately and outperforms the traditional method by 30%. Their frontier interpretable network prediction algorithm, RNet, demonstrates the best efficiency and precision. RNet positively impacts complex structure prediction, leading to a significant increase in the number of successful hits. This eliminates the need for time-consuming screening of thousands of potential binding candidates, making it an essential tool for efficient and practical research.

An additional limitation is to predict the binding structure accurately [17, 18]. In the article ‘Drug repositioning based on weighted local information augmented graph neural network’, Meng et al. [19] focused on drug repositioning, the strategy that aims at repurposing existing drugs for new therapeutic purposes, which is also a pivotal approach to accelerate drug discovery. They introduced a novel deep-learning method called DRAGNN for drug repositioning. This approach uses a graph neural network and includes weighted local information augmentation. Specifically, the authors first incorporate a graph attention mechanism to dynamically allocate attention coefficients to drug and disease heterogeneous nodes, enhancing the effectiveness of target node information collection. Their model demonstrates promising potential in predicting candidate drugs and drug combinations through molecular docking and network analysis experiments. In the article ‘Prediction of protein-ligand binding affinity via deep learning models’, Wang [20] provided us with a detailed overview of the current state-of-the-art computational methods, particularly deep learning-based models, that are being used to predict protein-ligand binding affinity. The author explained the fundamental principles of deep learning models in predicting the relationship between proteins and ligands. The author also examined the databases and input representations in this area. Based on the review, the author highlighted the potential challenges and future work required to predict the protein-ligand binding affinity accurately.

The articles in this special issue showcase diverse advanced techniques for studying proteins and RNAs. We hope these methods deliver substantial value and offer profound insights to the protein and RNA research community.

FUNDING

National Natural Science Foundation of China (grant no. 12175081); Fundamental Research Funds for the Central Universities (grant no. CCNU22QN004).

DATA AVAILABILITY

No new data were generated or analyzed in support of this manuscript.

Author Biographies

Haoquan Liu is a PhD student in the College of Physical Science and Technology at Central China Normal University. His research is centered around network science and deep learning, particularly in the field of biomolecule mechanisms and interactions.

Yunjie Zhao is a professor in the College of Physical Science and Technology at Central China Normal University. His research interests revolve around understanding the fundamental principles that underlie molecular structure and function. He focuses on designing biomolecules related to human diseases and developing computational tools to address challenges in health and technology.

References