Data-driven research on cell morphogenesis
Cells make decisions to move and form their own morphology by using their force. This project aims to elucidate this decision-making process from experimental data. Applying principles found in previous research (e.g., bistability from intermolecular interactions) to everything is mathematical-driven, not data-driven. We are conducting data-driven research on cell morphogenesis by carefully observing experimental data and considering physical quantities other than molecules as signals.

Data-driven research on biological tissue formation
We are trying to understand the formation of biological tissues by a variety of physical quantities. By observing biological data, we find that organisms are not robust to external stimuli, but rather resilient, able to recover from the effects of stimuli and achieve their goals. We are analyzing data and developing quantitative models for resilient somitogenesis, angiogenesis, and regulation of organ size.

Force estimation generated by cells using traction force microscopy images
Anything can be deformed only by force. Cells can also be deformed by forces. It is impossible to explain cellular deformation using only molecular information. In order to perform a data-driven study of cellular deformation, we need data on cellular forces. However, it is not possible to directly observe the cellular forces. We are developing a method to accurately estimate cellular forces in the framework of Bayesian statistics using data observed by traction force microscopy.

Transomics
The technology for simultaneous measurement of multiple molecules, called omics analysis, is being developed. This has led to the discovery of many molecules involved in diseases. However, we cannot understand the mechanisms of biological phenomena by simply listing these molecules. We aim to elucidate biological systems in a data-driven manner by using multiple omics data to construct a network consisting of many molecules (trans-omics analysis).

Real-time optimal control of cellular systems using control theory and machine learning
Cells regulate cellular functions such as migration, fate determination (proliferation and differentiation), and metabolism by sensing environmental information such as growth factors and hormones and driving intracellular signaling and gene expression. We use control theory and machine learning to study analytical algorithms for artificial real-time control of cellular functions by designing environmental information optimally. Our goal is to develop a data-driven cellular control system that integrates measurement, analysis, and control.

Diagnosis of disease by machine learning using human breath gas
Alcohol concentration in a driver’s blood is measured in the breath beacuse the breath component is extracted from the blood component in the alveoli. If this is the case, then it is possible to diagnose diseases using breath tests instead of invasive tests such as blood tests. We are working on this problem in collaboration with doctors and sensor development engineers. The advancement of non-invasive, simple tests with high accuracy will lead to preventive medicine.

Analysis and mathematical model of molecular transport and localization by intracellular actin wave
Supply is the most important factor in the management of a nation and war. Similarly, the transport of materials for cells is essential for the expression of functions. Energy does not simply fall everywhere. Looking at images of neurons, we can see that actin fibers are not only involved in skeletal formation, but also actin waves are involved in the transport of materials. We will build a quantitative mathematical model of the actin wave based on experimental data and elucidate its mathematical principles.

Development of methods to integrate varying data and individually observed data (dealing with system noise)
Cells respond in different ways even under the same conditions. In addition, multiple types of molecules and phenotypes cannot be observed simultaneously, but are recorded as individually observed data. Such ” variations” and “individual observations” are major problems in biology for analysis. They are not something that experimental researchers can solve. We aim to solve them with data-driven biology.

Development of semi-automatic quantification software for cell images (dealing with observation noise)
In human society, there are no perfect laws, but we have backup systems when problems occur. Similarly, when quantifying features from biological images, there is no fully automated algorithm. An automated algorithm that is specific to particular images is an “overlearning algorithm” and is not generalizable. We are developing “semi-automatic” software that can be easily backed up by users, rather than fully automated.

Network estimation using gene expression time series data
The interaction between genes during development has been studied previously. In recent years, time series of gene expression in developmental processes have become available and more information can be handled. We are attempting to estimate gene-gene interactions using the expression time series of many types of genes. In particular, we are trying to estimate gene networks in the development of the chick neural tube and mouse oocytes.