Article
Biochemical Research Methods
Jing Xu, Aidi Zhang, Fang Liu, Xiujun Zhang
Summary: This paper introduces an interpretable transformer-based method called STGRNS for inferring gene regulatory networks (GRNs) from scRNA-seq data. By using a gene expression motif technique, gene pairs are converted into contiguous sub-vectors for transformer encoding. The experimental results show that STGRNS outperforms other methods on various scRNA-seq data types and is more interpretable than black box deep learning methods.
Article
Biotechnology & Applied Microbiology
T. Lohoff, S. Ghazanfar, A. Missarova, N. Koulena, N. Pierson, J. A. Griffiths, E. S. Bardot, C. -H. L. Eng, R. C. V. Tyser, R. Argelaguet, C. Guibentif, S. Srinivas, J. Briscoe, B. D. Simons, A. -K. Hadjantonakis, B. Gottgens, W. Reik, J. Nichols, L. Cai, J. C. Marioni
Summary: Improved integration of spatial and single-cell transcriptomic data through the seqFISH method provides insights into mouse development, revealing cell types across the embryo and uncovering axes of cell differentiation that are not apparent from scRNA-seq data. This approach offers a high-resolution spatial map for studying cell fate decisions in complex tissues and development.
NATURE BIOTECHNOLOGY
(2022)
Article
Biochemistry & Molecular Biology
Xingyan Liu, Qunlun Shen, Shihua Zhang
Summary: Cross-species comparative analysis of scRNA-seq data is important for understanding the cellular diversity and evolutionary mechanisms. However, assigning cell identities for nonmodel species is challenging due to limited genome annotation and known biomarkers. In this study, we propose a heterogeneous graph neural network model, CAME, which improves cell-type characterization across distant species by using non-one-to-one homologous gene mapping.
Article
Computer Science, Interdisciplinary Applications
Andrea Tangherloni, Simone G. Riva, Brynelle Myers, Francesca M. Buffa, Paolo Cazzaniga
Summary: Single-cell RNA sequencing experiments are valuable for identifying different cell types. This study introduces a fully-automatic framework called MAGNETO, which constructs optimal marker panels for distinguishing desired cell populations. The results demonstrate that MAGNETO outperforms other methods in identifying the cell populations of interest.
JOURNAL OF BIOMEDICAL INFORMATICS
(2023)
Article
Biochemistry & Molecular Biology
Shanni Cao, Zhaohui He, Ruidong Chen, Yuting Luo, Liang-Yu Fu, Xinkai Zhou, Chao He, Wenhao Yan, Chen -Yu Zhang, Dijun Chen
Summary: Single-cell transcriptomics has been widely used in plant biological research, but analyzing single-cell transcriptomic data in plants is not easy. We introduce scPlant, a versatile framework that helps users explore plant single-cell atlases and provides various analysis and visualization tools.
PLANT COMMUNICATIONS
(2023)
Article
Biochemical Research Methods
Mengyuan Zhao, Wenying He, Jijun Tang, Quan Zou, Fei Guo
Summary: In this study, a deep learning framework called DGRNS is developed for inferring gene regulatory networks from single-cell transcriptomic data. It effectively identifies the relationships between gene pairs and overcomes the challenges of sparsity and noise in the data. The results show that DGRNS outperforms other methods and discovers novel regulatory relationships with high confidence, which require further research.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Zilong Zhang, Feifei Cui, Wei Su, Lijun Dou, Anqi Xu, Chen Cao, Quan Zou
Summary: In this study, a web-based scRNA-seq analysis tool called webSCST is introduced, which integrates well-organized spatial transcriptome sequencing datasets classified by species and organs. It provides a user-friendly interface for raw single-cell processing and popular integration methods, allowing users to obtain predicted spatial locations for each cell type by submitting their raw scRNA-seq data once.
Article
Biochemical Research Methods
Xishuang Dong, Shanta Chowdhury, Uboho Victor, Xiangfang Li, Lijun Qian
Summary: The researchers propose a semi-supervised learning model, SemiRNet, to identify cell types using unlabeled and limited labeled single-cell transcriptomic data. The model is based on recurrent convolutional neural networks and consists of shared, supervised, and unsupervised networks. Evaluation on two large-scale single-cell transcriptomic datasets shows that the proposed model achieves impressive performance by learning from a small number of labeled cells and a large number of unlabeled cells.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Xu Shan, Jinyu Chen, Kangning Dong, Wei Zhou, Shihua Zhang
Summary: Single-cell RNA sequencing (scRNA-seq) is limited in capturing the spatial organization of cells in a tissue, while spatially resolved transcriptomics technologies (ST) have been developed to address this issue. However, current ST technologies are inefficient at single-cell resolution. In this study, we introduced a partial least squares-based method (spatial modular patterns [SpaMOD]) to integrate scRNA-seq and ST data, and identify cell-spot comodules for deciphering tissue spatial patterns.
JOURNAL OF COMPUTATIONAL BIOLOGY
(2022)
Article
Biochemistry & Molecular Biology
Junil Kim, Simon T. Jakobsen, Kedar N. Natarajan, Kyoung-Jae Won
Summary: Accurate prediction of gene regulatory rules is crucial for understanding cellular processes. The novel approach TENET utilizes transfer entropy to reconstruct gene regulatory networks from single cell RNA sequencing data, outperforming other reconstructors in identifying key regulators. Results show that known target genes have significantly higher TE values and genes predicted by TENET to have higher TE are more influenced by their regulator's perturbation.
NUCLEIC ACIDS RESEARCH
(2021)
Article
Biochemical Research Methods
Leah L. Weber, Palash Sashittal, Mohammed El-Kebir
Summary: The study introduces a standalone method, doubletD, for detecting doublets in scDNA-seq data, achieving high performance through a simple maximum likelihood approach that outperforms current methods. Applying doubletD in scDNA-seq analysis pipelines can reduce complexity and improve result accuracy.
Article
Biochemical Research Methods
Zhenhua Yu, Fang Du
Summary: Single-cell DNA sequencing allows for high-resolution analysis of intra-tumor heterogeneity. Existing methods for phylogenetic inference from scDNA-seq data perform well on small datasets, but are computationally inefficient and less accurate on large datasets. In this study, we introduce a new software called AMC that accurately clusters mutations, improving the efficiency of phylogenetic inference.
Article
Multidisciplinary Sciences
Aleksandr Ianevski, Anil K. Giri, Tero Aittokallio
Summary: We developed a computational platform, ScType, for automated and fast cell-type identification using scRNA-seq data and a comprehensive cell marker database. ScType provides unbiased and accurate cell type annotations by ensuring the specificity of marker genes across cell clusters and types. It can also distinguish between healthy and malignant cell populations based on single-nucleotide variant calling.
NATURE COMMUNICATIONS
(2022)
Article
Biochemistry & Molecular Biology
Brendan F. Miller, Dhananjay Bambah-Mukku, Catherine Dulac, Xiaowei Zhuang, Jean Fan
Summary: The computational framework MERINGUE is developed for spatially resolved transcriptomic data analysis, enabling cell clustering and identification of gene expression patterns in 2D and 3D. This spatial analysis method is expected to enhance our understanding of the interplay between cell state and spatial organization in tissue development and disease.
Article
Multidisciplinary Sciences
Alok K. Maity, Andrew E. Teschendorff
Summary: This study introduces a differential abundance testing paradigm called ELVAR, which uses cell attribute aware clustering to infer differentially enriched communities within the single-cell manifold. By benchmarking ELVAR against other algorithms using simulated and real datasets, the authors demonstrate that ELVAR improves the sensitivity to detect cell-type composition shifts in relation to aging, precancerous states, and Covid-19 phenotypes. Leveraging cell attribute information helps denoise single-cell data, avoid batch correction, and retrieve more robust cell states for subsequent differential abundance testing.
NATURE COMMUNICATIONS
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Ji-Young Kim, Dongkyu Oh, Kiyoung Sung, Hongyoon Choi, Jin Chul Paeng, Gi Jeong Cheon, Keon Wook Kang, Dong Young Lee, Dong Soo Lee
Summary: This study analyzed the impact of deep learning-based one-step amyloid burden estimation system on inter-reader agreement and confidence of reading in clinical routine amyloid PET reading. The results showed that the deep learning system improved inter-reader agreement and increased confidence in visual interpretation.
EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING
(2021)
Article
Biochemistry & Molecular Biology
Sungwoo Bae, Hongyoon Choi, Dong Soo Lee
Summary: The study presents a new method, SPADE, for identifying important genes associated with morphological contexts by combining spatial transcriptomic data with coregistered images. SPADE uses deep learning-derived image patterns to extract morphological context markers from spatially resolved gene expression data. Molecular profiles correlated with the image latent features can be identified, allowing for further analysis of extracted genes to discover functional terms and extract clusters maintaining morphological contexts. The approach is demonstrated on spatial transcriptomic data from different tissues, platforms, and image types, showing its capability of obtaining unbiased image-integrated gene expression trends.
NUCLEIC ACIDS RESEARCH
(2021)
Article
Oncology
Seong-Woo Bae, Felix Berlth, Kyoung-Yun Jeong, Ji-Hyeon Park, Jong-Ho Choi, Shin-Hoo Park, Yun-Suhk Suh, Seong-Ho Kong, Do-Joong Park, Hyuk-Joon Lee, Charles Lee, Jong-Il Kim, Hyewon Youn, Hongyoon Choi, Gi Jeong Cheon, Keon Wook Kang, Han-Kwang Yang
Summary: This study developed a genetic signature, PETscore, to predict the metabolic activity of GC and identified five genes associated with [F-18]FDG uptake in GC. The PETscore was validated using human GC data and revealed associations between glucose uptake, tumor mutational burden, and genomic alterations in GC.
Article
Dentistry, Oral Surgery & Medicine
Jung Hwan Jo, Sungwoo Bae, Joonhyung Gil, Dongkyu Oh, Seoeun Park, Gi Jeong Cheon, Ji Woon Park
Summary: This study aimed to verify the association between initial bone scintigraphy results and long-term TMJ DJD prognosis. The results showed that initial bone scintigraphy did not have sufficiently close associations with long-term TMJ DJD prognosis, suggesting the need for further studies to develop prognostic indices that combine clinical and imaging contents.
JOURNAL OF ORAL REHABILITATION
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Donghwi Hwang, Seung Kwan Kang, Kyeong Yun Kim, Hongyoon Choi, Jae Sung Lee
Summary: This study compares two approaches utilizing PET data and a convolution neural network (CNN) for attenuation correction of annihilation photons in PET. Results indicate that the scatter estimation from mu-CNNNAC is valid despite less accurate bone structures. The combination of mu-CNNMLAA+NAC provides the best results in recovering fine bone structures.
EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Reeree Lee, Hongyoon Choi, Kwang-Yeol Park, Jeong-Min Kim, Ju Won Seok
Summary: Post-stroke cognitive impairment is common and affects many stroke survivors. A deep-learning-based signature using PET has shown promising results in objectively evaluating cognitive decline in stroke patients, with potential clinical applications.
EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING
(2022)
Article
Oncology
Hongyoon Choi, Kwon Joong Na
Summary: This study comprehensively investigated the glucose metabolic features of the TME at the single-cell level and found that differently expressed GLUTs could serve as potential targets for tumor immune status and adjunctive treatments for immunotherapy.
FRONTIERS IN ONCOLOGY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Wonseok Whi, Hongyoon Choi, Jin Chul Paeng, Gi Jeong Cheon, Keon Wook Kang, Dong Soo Lee
Summary: In this study, a fully automated quantitative interpretation pipeline of brain volume from oncology PET images was developed using deep learning, which successfully identified brain volume and detected metastatic lesions. The approach achieved a high accuracy rate in identifying brain abnormalities and has potential for quantitative analysis in oncologic PET studies.
Article
Radiology, Nuclear Medicine & Medical Imaging
Dongkyu Oh, Hongyoon Choi, Jin Chul Paeng, Keon Wook Kang, Gi Jeong Cheon
Summary: The uptake of (68) Ga-DOTA-TOC in the pancreas uncinate process is negatively correlated with blood glucose levels. This suggests that glycemia may affect the physiologic uptake of (68) Ga-DOTA-TOC.
NUCLEAR MEDICINE AND MOLECULAR IMAGING
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Jimin Hong, Seung Kwan Kang, Ian Alberts, Jiaying Lu, Raphael Sznitman, Jae Sung Lee, Axel Rominger, Hongyoon Choi, Kuangyu Shi
Summary: This study used a data-driven approach to identify the tau trajectory and quantify tau progression through a continuous latent space learned by VAE. The inferred tau trajectory in line with Braak staging showed tau first deposits in the parahippocampal and amygdala before spreading to other specific brain regions.
EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING
(2022)
Article
Oncology
Hyunjong Lee, Hongyoon Choi
Summary: This study constructed a pseudotime trajectory model using clinical and radiogenomic data of lung adenocarcinoma (LUAD) patients, and found that the pseudotime trajectories corresponded to clinical stages. The study also identified dynamic changes in molecular features and immune cell activity along the pseudotime trajectory.
FRONTIERS IN ONCOLOGY
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Junyoung Park, Seung Kwan Kang, Donghwi Hwang, Hongyoon Choi, Seunggyun Ha, Jong Mo Seo, Jae Seon Eo, Jae Sung Lee
Summary: The study proposes a two-stage U-Net architecture to enhance the performance of lung cancer segmentation using [F-18]FDG PET/CT. The method outperforms the conventional one-stage approach and reduces the time and effort required for accurate lung cancer segmentation.
NUCLEAR MEDICINE AND MOLECULAR IMAGING
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Jinyeong Choi, Azmal Sarker, Hongyoon Choi, Dong Soo Lee, Hyung-Jun Im
Summary: This study aimed to stratify the prognosis of lung adenocarcinoma patients using F-18-FDG PET parameters and immune cell scores. Results showed that different immune cell scores and CYT were associated with prognosis, and high TLR and TFH were predictive of overall survival independently in lung adenocarcinoma patients.