The University of Osaka · Engineering
Professor Takashi Washio's research lab specializes in the integration of machine learning and experimental chemistry to accelerate discovery in synthetic organic chemistry and materials science. The lab focuses on developing intelligent screening strategies—particularly using Bayesian optimization and active learning—for complex multivariate reactions such as electrochemical transformations, organocatalysis, and flow chemistry. A key research direction involves enabling precise, single-molecule detection and analysis through machine learning-enhanced signal processing. The lab also pioneers data-driven discovery systems that extract meaningful chemical relationships from experimental data with high robustness and interpretability.
Figures are computed from collected data and may differ slightly.
The need for mining structured data has increased in the past few years. One of the best studied data structures in computer science and discrete mathematics are graphs. It can therefore be no surprise that graph based data mining has become quite popular in the last few years.This article introduces the theoretical basis of graph based data mining and surveys the state of the art of graph-based data mining. Brief descriptions of some representative approaches are provided as well.
A highly atom-economical enantioselective organocatalyzed Rauhut-Currier and [3+2] annulation sequence has been established by using a flow system. Suitable flow conditions were explored through reaction screening of multiple parameters using machine learning. Eventually, functionalized chiral spirooxindole analogues were obtained in high yield with good ee as a single diastereomer within one minute.
When conducting a single-molecule measurement, data are analyzed using the histogram of a measured physical quantity in which a single dataset contains information about a specific single molecule. Oftentimes, the histogram consisting of only specific single-molecule information excludes the input from other information sources. In other words, despite measuring the single molecule during analysis, we miss many of the properties of that single molecule. Herein, we have successfully identified a
Multiparameter screening of reductive carboxylation in an electrochemical flow microreactor was performed using a Bayesian optimization (BO) strategy. The developed algorithm features a constraint on passed charge for the electrochemical reaction, which led to suitable conditions being instantaneously found for the desired reaction. Analysis of the BO-suggested conditions underscored the physicochemical validity.
SDS is a discovery system from numeric measurement data. It outperforms the existing systems in every aspect of search e ciency, noise tolerancy, credibility of the resulting equations and complexity of the target system that it can handle. The power of SDS comes from the use of the scale-types of the measurement data and mathematical property of identity by which to constrain the admissible solutions. Its algorithm is described with a complex working example and the performance comparison with
A highly efficient synthesis of α-ketiminophosphonates has been established for the electrochemical oxidation of α-amino phosphonates with the utilization of machine-learning-assisted simultaneous multiparameter screening.
The variability of bioparticles remains a key barrier to realizing the competent potential of nanoscale detection into a digital diagnosis of an extraneous object that causes an infectious disease. Here, we report label-free virus identification based on machine-learning classification. Single virus particles were detected using nanopores, and resistive-pulse waveforms were analyzed multilaterally using artificial intelligence. In the discrimination, over 99% accuracy for five different virus sp
Most conventional law equation discovery systems suchasBACON require experimental environments to acquire their necessary data. The mathematical techniques such as linear system identification and neural network fitting presume the classes of equations to model given observed data sets. The study reported in this paper proposes a novel method to discover an admissible model equation from a given set of observed data, while the equation is ensured to reflect first principles governing the objecti
Open papers in the app to read, cite, and organize with AI.
