DrugMatrix® The DrugMatrix database is the foundation for the development of ToxFX and the analysis of submitted compounds. Over the last several years the toxicology community has significantly increased the use of genome-wide gene expression profiling to dissect the mechanisms behind chemical toxicity and to increase the accuracy and sensitivity of toxicity testing. However, in order to appropriately supplement the classical panel of toxicity markers, gene expression-based biomarkers require the perspective of a large-scale contextual database in order to draw meaningful conclusions from novel and disconnected endpoints. Only after specific gene expression responses to diverse physiological, pharmacological, toxicological and pathological conditions have been observed multiple times across multiple causative compounds with diverse chemistry can the significance of that event be understood and genomic biomarkers faithfully applied. A large database of expression profiles combined with measured pharmacological and pathological responses to treatment with hundreds of different drugs and drug-like compounds, and the associated gene expression profiles is therefore essential for the development of gene expression biomarkers. The availability of such a reference database expedites the implementation and acceptance of genomic data in drug development and with regulatory agencies. DrugMatrix is an integrated informatics system that combines ready data access with multiple dimensions of data from animals treated with reference compounds in vivo as well as cells treated in vitro with relevant chemical compounds. The primary dimensions of data integrated by DrugMatrix are those drawn from new experimental results—gene expression profiles, molecular pharmacology profiles from a panel of 130 enzyme binding, receptor, and ion channel assays and blood chemistry and histopathology readings—as well as extensive hand-curated literature profiles extracted from multiple sources.
Drug Signature® Library The application of Drug Signatures to the analysis of submitted compounds is a key component of the ToxFX product. The complexity of gene expression analysis and data interpretation from massive datasets like DrugMatrix demands the development and application of clear and statistically validated analysis and visualization tools. Iconix applies a variety of bioinformatics tools such as unsupervised clustering, principal components analysis, pathway analysis, and various ANOVA (analysis of variance) techniques commonly used for classifying and characterizing compound toxicity and mechanism of action. Most importantly, Iconix has developed a proprietary set of Linear Classifier algorithms to mine the vast DrugMatrix dataset and identify Drug Signatures—sets of genes (i.e. biomarkers) whose patterns of expression are predictive of classical as well as novel pathological and pharmacological endpoints—in collaboration with academic researchers at the University of California Berkeley. Drug Signatures have been statistically validated and have been demonstrated to diagnose and predict compound activity and toxicity in preclinical models.
The ToxFX™ Array Design The ToxFX array provides a comprehensive analysis of gene expression changes at a lower price point than analysis on whole genome arrays. In order to discover patterns of differential gene expression that are associated with measurable biological endpoints, the scientists at Iconix submitted the DrugMatrix database to a systematic mining effort and attempted to derive validated Drug Signatures for every measured endpoint. These endpoints included the pathologies defined by the clinical measurements and histopathological observations described above. We also separated groups of related compounds based on structural similarity, pharmacological profile or literature annotations. Several hundred thoroughly cross-validated toxicology and pharmacology signatures, composed of an average of 47 genes each, were identified. We noted that some of the signature genes are present in multiple signatures and contribute disproportionately to the classification potential across all endpoints. The fact that all classifiable endpoints in this large dataset converge on a relatively limited number of genes reflects the biological reality that there are a small number of key processes involved in many toxicological and pharmacological responses to xenobiotic insults. In addition, many genes respond in a coordinated fashion to a given stimulus, resulting in significant redundancy in genes that respond to a particular xenobiotic insult. The identification of the key, non-redundant classifier genes from the DrugMatrix dataset formed the basis of the ToxFx chip design. In addition, Iconix scientists have selected 23 pathways from the 135 available in DrugMatrix that are of greatest relevance to rodent preclinical pathology and toxicology. Any genes from these pathways that do not already appear in a ToxFX Drug Signature were also added to the array. There are a total of 2073 probe sets present on the ToxFx chip. This represents the non-redundant union of the heart, kidney and liver classifier genes, non-redundant genes from the toxicologically important pathways. And a set of genes not present in either ToxFX Drug Signatures or pathways chosen on the basis of their relevance to toxicology. These genes include a comprehensive set of phase I (P450) and Phase II xenobiotic metabolism genes, as well as key genes involved in stress-signaling and adaptive response, cell cycle control, DNA repair, inflammation and tissue repair.
