Biology subjects

Terayama, K.

Publications and source records attributed to Terayama, K..

6 recordsLinked to original sources

bioRxiv2026

Resolution-standardized evaluation of ligand atomic coordinates in crystallographic structures using machine learning

Accurate assessment of ligand coordinate-density consistency across different resolutions remains challenging in macromolecular crystallography. We introduce the atomic Box Correlation Coefficient (aBCC), an atom-level metric for evaluating the consistency between ligand atomic coordinates and electron density in a resolution-standardized framework. To predict aBCC values from electron-density maps, we developed QAEmap, a machine-learning model based on three-dimensional convolutional neural networks (3D-CNNs). The model was trained using Fourier-truncated electron-density maps and corresponding ligand coordinates generated from high-resolution structures in the Protein Data Bank. It was evaluated using both Fourier-truncated electron-density maps and experimentally determined PDB structures. was evaluated using both Fourier-truncated electron-density maps and experimentally determined PDB structures.The prediction accuracy gradually decreased with decreasing resolution, but remained reliable up to [~]3.5 [A]. These results demonstrate that aBCC enables resolution-standardized atom-wise evaluation of coordinate-density consistency across different resolutions and provide a foundation for further development and refinement of machine learning-based coordinate validation. SynopsisWe introduce the atomic box correlation coefficient (aBCC), a machine learning-based metric for the resolution-standardized atom-level evaluation of ligand coordinate-density consistency in crystallographic structures. aBCC provides a common framework for assessing and communicating the local coordinate reliability between structural biologists and researchers in structure-based drug discovery.

molecular biology↗

bioRxiv2025

Assessing the Performance of LLMs in Multimodal Information Extraction for Biological Research: A Case Study on LLPS

Advances in experimental techniques have expanded the volume of biological data. This has increased the demand for structured information extraction from papers, with large language models (LLMs) considered promising. However, challenges remain, including limited validation in biology and unclear applicability to multimodal tasks that integrate text with domain-specific figures, such as microscopic images and scatter plots. Here, we developed a multimodal LLM (MLLM)-based workflow to extract the experimental conditions and phase status from the text and figures of experimental papers on liquid-liquid phase separation and validated the effect of various inputs, prompts, and MLLM types. As a result, the Gemini 2.5 Pro-based extraction achieved an F1-score of 0.847 by processing each figure as a processing unit and inputting domain-specific prompts reflecting manual extraction guidance. This study demonstrates the potential and limitations of MLLMs for extracting biological information and provides insights for advancing multimodal approaches in biology.

bioinformatics↗

bioRxiv2025

Data-Driven Feed Optimization for Sustainable Aquaculture through Integrated Omics Analysis and Two-stage Bayesian optimization

In aquaculture and livestock farming, developing feed that promotes efficient growth while minimizing environmental impact is crucial for sustainability. We developed a data-driven feed optimization system (DFOS) that integrates omics analysis with machine learning-based optimization technologies. First, DFOS employed Bayesian optimization (BO) to identify the optimal proportions of proteins and lipids, which are crucial feed components. Simultaneously, omics analysis was conducted to assess the physiological response of the targets to a given feed and identify key dietary components essential for growth. Next, BO was used to efficiently modify the combination and proportion of the specific additives identified in the first stage. We demonstrated the efficacy of DFOS by applying it to develop a new feed for the high-demand leopard coral grouper (Plectropomus leopardus) in Southeast Asia. This system provides an efficient data-driven framework for feed optimization across aquaculture and livestock, contributing significantly to more sustainable and productive farming practices.

ecology↗

bioRxiv2024

cryoMDM; Molecular simulation Driven structural Matching Approach to Estimate Variety of Atomic Three-dimensional Model Based on Noisy cryoEM Single Particle Images.

Understanding the structure of biomolecules is essential for comprehending life phenomena. In recent years, cryo-electron microscopy (cryoEM) single-particle structure analysis has made remarkable advancements. The single-particle images obtained from cryoEM measurements have a poor signal-to-noise ratio, and existing methods rely on multi-image analysis, which averages multiple measurement data to achieve high-resolution structures. However, this multi-image approach statistically processes several measurement images, causing the unique structural features of individual data to be lost. To overcome this inherent challenge, we developed a molecular simulation-driven structural matching method, cryoMDM, which identifies the plausible structure for cryoEM two-dimensional images from a vast number of three-dimensional (3D) structural model candidates generated through molecular simulations. The innovative aspect of this method is that it enables the estimation of 3D atomic models from a single-particle image, going beyond merely generating a 3D electron density map. By linking each cryoEM particle image to a 3D structure in the simulation space, we can directly connect cryoEM measurement data to continuous structural changes in biomolecules. Applying this method to the spike protein derived from SARS-CoV-2, we successfully captured various intermediate structures of the spike protein, revealing critical information about the early stages of viral entry.

biophysics↗

bioRxiv2023

Predicting condensate formation of protein and RNA under various environmental conditions

MotivationLiquid-liquid phase separation (LLPS) by biomolecules plays a central role in various biological phenomena and has garnered significant attention. The behavior of LLPS is strongly influenced by the characteristics of the RNAs and environmental factors such as pH and temperature, as well as the properties of the proteins. Recently, several databases of biomolecules associated with LLPS have been established, and prediction models of LLPS-related phenomena have been explored, leveraging these databases. However, a prediction model that concurrently considers proteins, RNAs, and experimental conditions has not been developed due to the limited information available from individual experiments in public databases. ResultsTo address this challenge, we have built a new dataset called RNAPSEC, which serves each individual experiment as a data point. This dataset was accomplished by manually collecting data from public literature. Utilizing RNAPSEC, we developed two distinct models that consider a protein, RNA, and experimental conditions. The first model can predict the LLPS behavior of a protein and RNA under specific conditions. The second model can predict the required conditions for a given protein and RNA to undergo LLPS. RNAPSEC and these prediction models are expected to accelerate our understanding of the roles of proteins, RNAs, and environmental factors in LLPS. AvailabilityThe codes for the prediction models and RNAPSEC are available at https://github.com/ycu-iil/RNAPSEC. Contactterayama@yokohama-cu.ac.jp

bioinformatics↗

bioRxiv2020

Extraction of Protein Dynamics Information Hidden in Cryo-EM Map Using Deep Learning

Technical breakthroughs in cryogenic electron microscopy (cryo-EM)-based single-particle analysis have enabled the structures of numerous proteins to be solved at atomic or near-atomic resolutions, including extremely large macromolecules whose structures could not be solved by conventional techniques. Determining the dynamics properties of these macromolecules, based on their solved structures, can further improve our understanding of their functional mechanisms. However, such analysis is often hampered by the large molecular size and complex structural assembly, making both experimental and computational approaches to determine dynamics properties challenging. Here, we report a deep learning-based approach, DEFMap, to extract the dynamics information "hidden" in cryo-EM density maps. By relying only on cryo-EM maps, DEFMap successfully provided dynamics information equivalent to that determined from molecular dynamics (MD) simulations and experimental approaches at the atomic and residue levels. Additionally, DEFMap could detect dynamics changes associated with molecular recognition and the accompanying allosteric conformational stabilizations, which trigger various biological events such as signal transduction and enzyme catalysis. This approach will provide new insights into the functional mechanisms of biological molecules, accelerating modern molecular biology researches. Furthermore, this advanced strategy combining experimental data, deep learning approaches, and MD simulations would open a new multidisciplinary science area

molecular biology↗