Host: The Japanese Society of Toxicology
‘Percellome’ database is providing a unique “per cell” readout in mRNA copy number on various organs of mice for more than 100 chemicals (NIHS). Such knowledge makes possible to use dataset without data pre-elimination and compare profiles across experiments. This opens a possibility for the mathematical representation of biological information in a manner treating a cell as an open system, with no molecule ignored.
In collaboration with Kanno group we constructed A Geometric Clustering Tool (AGCT) and a comparative genomic analysis tool on human-mouse-rat (SHOE) which goal is to divide an expression profile by unsupervised learning mathematical technique in order to elucidate gene clusters and find transcription regulation network controlling genes. Analysis was held on two toxic congeners TCDD and TCDF, with 0.1 toxicity equivalency factor. They are known to cause birth defects, immunotoxicity and cancer through the activation of Ahr receptor pathway in mice, whilst Ahr receptor is a mediator in the course of growth, development and differentiation.
Data of 20,000 probes per cell followed: replicates normalization, circadian effect subtraction, representation by PCA and/or spectral manifold, unsupervised clustering and sorting by validity. The analysis resulted in 498 (3,117 probes) and 369 (3,771 probes) sensible clusters for TCDD and TCDF, respectively, meeting the biological expectations. Clustering and transcription regulation results can be simulated on CellDesigner (SBI) and interactively annotated on Payao (Matsuoka, SBI).
Since this year Percellome database is a member of The Garuda Alliance Common Platform, providing the research community with the consistent analysis workflow and biological databases.