Annual Report 2024
Laboratory of Computational Life Science
Yasuhiro Kojima, Kazuya Nishimura, Naoko Takagi
Introduction
Recent advancements in single-cell and spatial omics analysis have significantly increased the complexity of omics data, particularly transcriptomics. The Computational Life Science Unit is developing computational methodologies to efficiently extract knowledge from these comprehensive gene expression profiles for single cells and spatial niches within tissues.
The Team and What We Do
We are developing computational methodologies by combining mathematical models and deep learning. In this fiscal year, we developed a new method for learning microenvironmental states using deep learning and validated its performance on both real and simulated datasets. In addition, we developed a deep learning model for integrative analysis of pathology images and spatial transcriptomics data.
Research Activities
DeepKINET: A deep generative model for estimating single-cell RNA splicing and degradation rates (Genome Biology, September 6, 2024)
DeepKINET is a deep generative model that enables estimation of splicing and degradation rates at the cell and gene levels from single-cell RNA-seq data. Unlike conventional methods that assume "constant rates," it allows analysis of the diversity of regulation by RNA-binding proteins and splicing factors. Through application to datasets from the forebrain, breast cancer, and myelodysplastic syndrome, this work contributes to understanding post-transcriptional regulation and its application to disease research.
LineageVAE: A generative model for reconstructing unobserved ancestral cell states and transcriptional profiles (Bioinformatics, August 22, 2024)
LineageVAE is a variational autoencoder model that integrates scRNA-seq data with lineage barcode information to estimate unobserved ancestral cell states and past expression profiles. Using hematopoiesis and cell reprogramming datasets, the model successfully reconstructed past states consistent with known marker expression. This method provides a powerful approach to deepen our understanding of developmental and differentiation processes, as well as cell lineage dynamics.
Education
We have hosted 2 specially appointed research fellows, 1 specially appointed research assistant, 2 graduate students from partner universities, and 4 voluntary trainees. Guidance has been provided through individual meetings, progress report sessions, and journal club.
Future Prospects
Thus far, we have elucidated the biological significance and mechanisms of cellular heterogeneity by linking single-cell-level diversity to various aspects of organismal systems, such as spatial localization, splicing, and temporal dynamics. At the same time, multicellular tissues, including tumors, are dynamic systems shaped by complex interactions among multiple cells. Therefore, our future research will focus on uncovering the diversity of the tissue microenvironment and the tissue itself, as well as the underlying systems that generate such diversity.
List of papers published in 2024
Journal
1. Sato T, Sugiyama D, Koseki J, Kojima Y, Hattori S, Sone K, Nishinakamura H, Ishikawa T, Ishikawa Y, Kato T, Kiyoi H, Nishikawa H. Sustained inhibition of CSF1R signaling augments antitumor immunity through inhibiting tumor-associated macrophages. JCI insight, 10:e178146, 2025
2. Hashimoto M, Kojima Y, Sakamoto T, Ozato Y, Nakano Y, Abe T, Hosoda K, Saito H, Higuchi S, Hisamatsu Y, Toshima T, Yonemura Y, Masuda T, Hata T, Nagayama S, Kagawa K, Goto Y, Utou M, Gamachi A, Imamura K, Kuze Y, Zenkoh J, Suzuki A, Takahashi K, Niida A, Hirose H, Hayashi S, Koseki J, Fukuchi S, Murakami K, Yoshizumi T, Kadomatsu K, Tobo T, Oda Y, Uemura M, Eguchi H, Doki Y, Mori M, Oshima M, Shibata T, Suzuki Y, Shimamura T, Mimori K. Spatial and single-cell colocalisation analysis reveals MDK-mediated immunosuppressive environment with regulatory T cells in colorectal carcinogenesis. EBioMedicine, 103:105102, 2024
3. Majima K, Kojima Y, Minoura K, Abe K, Hirose H, Shimamura T. LineageVAE: reconstructing historical cell states and transcriptomes toward unobserved progenitors. Bioinformatics (Oxford, England), 40:btae520, 2024
4. Mizukoshi C, Kojima Y, Nomura S, Hayashi S, Abe K, Shimamura T. DeepKINET: a deep generative model for estimating single-cell RNA splicing and degradation rates. Genome biology, 25:229, 2024