Abstract
A living organism is a self-contained system driven by a variety of internal signals governed by receptors to interact with the environment. Although signals are measurable by omics technologies, the entire system is still too complex to decipher completely. One way to approach the system through accumulated measurements is to apply the approach for the “Big Data”.
Since 2003, we have been developing the Percellome databases consisting of mouse transcriptome data exposed to various chemicals measured by Affymetrix GeneChip expression arrays. Crucial points to obtain high-quality, reliable data are managing circadian rhythm of the mice, and the applying the “Percellome” normalization method (BMC Genomics. 2006 Mar 29;7:64.). At present, the Percellome database has comprehensive transcriptome data of liver and other organs on 151 chemicals by oral or inhalation route of exposure. Prediction of toxicity and mining of new mechanistic hypothesis of toxicity can be performed through a comparative analysis of “gene lists” made of significantly altered genes by the chemical administration.
Those new hypotheses are often verified by literature-based databases. Applying AI to such process is one of our approaches to integrate AI methods to our project. Another approach includes introduction of AI image recognition technology to verify the biological significance of the gene expression data. Any AI improvements will be reflected in the Percellome open database service soon. The goal is to evaluate or predicte the toxicity of new chemicals from short and small-scale animal bioassays.
