Chemogenomic profiling on a genome-wide scale using reverse-engineered gene networks

A major challenge in drug discovery is to distinguish the molecular targets of a bioactive compound from the hundreds to thousands of additional gene products that respond indirectly to changes in the activity of the targets. Here, we present an integrated computational-experimental approach for computing the likelihood that gene products and associated pathways are targets of a compound. This is achieved by filtering the mRNA expression profile of compound-exposed cells using a reverse-engineered model of the cell's gene regulatory network. We apply the method to a set of 515 whole-genome yeast expression profiles resulting from a variety of treatments (compounds, knockouts and induced expression), and correctly enrich for the known targets and associated pathways in the majority of compounds examined. We demonstrate our approach with PTSB, a growth inhibitory compound with a previously unknown mode of action, by predicting and validating thioredoxin and thioredoxin reductase as its target.     

Do predictive signatures really predict response to cancer chemotherapy?

Predictive signatures are gene expression profiles that should predict the response of tumors to chemotherapy in patients. Such signatures have been derived from the response of tumor cell lines to chemotherapy, but their usefulness in patients remains controversial, as the most spectacular published signatures are based on unreliable data. We discuss why it is difficult to derive meaningful predictive signatures from cell line panels and we argue that it is implausible that fully predictive signatures can be obtained for classical chemotherapy from oligo-based gene expression arrays. One reason is that resistance to chemotherapy can be caused by alterations in (the expression of) a single gene. We do not expect that such subtle alterations will be reliably picked up by standard gene expression profiling. We delineate alternative approaches that should be able to yield predictive markers that can be used for optimizing patient treatment.     

Methylation of cancer-stem-cell-associated Wnt target genes predicts poor prognosis in colorectal cancer patients

Gene signatures derived from cancer stem cells (CSCs) predict tumor recurrence for many forms of cancer. Here, we derived a gene signature for colorectal CSCs defined by high Wnt signaling activity, which in agreement with previous observations predicts poor prognosis. Surprisingly, however, we found that elevated expression of Wnt targets was actually associated with good prognosis, while patient tumors with low expression of Wnt target genes segregated with immature stem cell signatures. We discovered that several Wnt target genes, including ASCL2 and LGR5, become silenced by CpG island methylation during progression of tumorigenesis, and that their re-expression was associated with reduced tumor growth. Taken together, our data show that promoter methylation of Wnt target genes is a strong predictor for recurrence of colorectal cancer, and suggest that CSC gene signatures, rather than reflecting CSC numbers, may reflect differentiation status of the malignant tissue.     

Cross-Platform Toxicogenomics for the Prediction of Non-Genotoxic Hepatocarcinogenesis in Rat

In the area of omics profiling in toxicology, i.e. toxicogenomics, characteristic molecular profiles have previously been incorporated into prediction models for early assessment of a carcinogenic potential and mechanism-based classification of compounds. Traditionally, the biomarker signatures used for model construction were derived from individual high-throughput techniques, such as microarrays designed for monitoring global mRNA expression. In this study, we built predictive models by integrating omics data across complementary microarray platforms and introduced new concepts for modeling of pathway alterations and molecular interactions between multiple biological layers. We trained and evaluated diverse machine learning-based models, differing in the incorporated features and learning algorithms on a cross-omics dataset encompassing mRNA, miRNA, and protein expression profiles obtained from rat liver samples treated with a heterogeneous set of substances. Most of these compounds could be unambiguously classified as genotoxic carcinogens, non-genotoxic carcinogens, or non-hepatocarcinogens based on evidence from published studies. Since mixed characteristics were reported for the compounds Cyproterone acetate, Thioacetamide, and Wy-14643, we reclassified these compounds as either genotoxic or non-genotoxic carcinogens based on their molecular profiles. Evaluating our toxicogenomics models in a repeated external cross-validation procedure, we demonstrated that the prediction accuracy of our models could be increased by joining the biomarker signatures across multiple biological layers and by adding complex features derived from cross-platform integration of the omics data. Furthermore, we found that adding these features resulted in a better separation of the compound classes and a more confident reclassification of the three undefined compounds as non-genotoxic carcinogens.     

Drug response prediction by ensemble learning and drug-induced gene expression signatures

Chemotherapeutic response of cancer cells to a given compound is one of the most fundamental information one requires to design anti-cancer drugs. Recently, considerable amount of drug-induced gene expression data has become publicly available, in addition to cytotoxicity databases. These large sets of data provided an opportunity to apply machine learning methods to predict drug activity. However, due to the complexity of cancer drug mechanisms, none of the existing methods is perfect. In this paper, we propose a novel ensemble learning method to predict drug response. In addition, we attempt to use the drug screen data together with two novel signatures produced from the drug-induced gene expression profiles of cancer cell lines. Finally, we evaluate predictions by in vitro experiments in addition to the tests on data sets. The predictions of the methods, the signatures and the software are available from http://mtan.etu.edu.tr/drug-response-prediction/.