However, even cells with same genetic information can respond differently to the same external stimuli, called cellular heterogeneity. The sources of such heterogeneity and its relationship with the signaling system have remained a challenge, as intermediate processes of the signaling system are impossible to fully observe with current experimental technology. To reveal the sources of this heterogeneity, Professor Kim's research team developed a machine learning methodology using artificial neural network structures called Density Physics-informed neural networks (Density-PINNs). Density-PINNs use the observable time-series data of cells' responses to external stimuli to inversely estimate information about the signaling system.

 

By applying Density-PINNs to actual experimental data of antibiotic responses of bacterial cells (Escherichia coli), the research team found that a parallel structure of the signaling system can reduce heterogeneity among cells. Dr. JO Hyeontae and Dr. HONG Hyukpyo participated as co-first authors in this research, which was published in the international journal Patterns (Impact Factor 6.5), a sister journal of Cell.