BISON: Bio-Interface for the Semi-global analysis Of Network patterns

Predicting response to preoperative chemotherapy agents by identifying drug action on modeled microRNA regulation networks

Identifying patients most responsive to specific chemotherapy agents in neoadjuvant settings can help to maximize the benefits of treatment and minimize unnecessary side effects. Metagene approaches that predict response based on gene expression signatures derived from an associative analysis of clinical data can identify chance associations caused by the heterogeneity of a tumor, leading to reproducibility issues in independent validations. In this study, to incorporate information from drug mechanisms of action, we explore the potential of microRNA regulation networks as a new feature space for identifying predictive markers. We introduce a measure we term the CoMi (Context-specific-miRNA-regulation) pattern to represent a descriptive feature of the miRNA regulation network in the transcriptome. We examine whether the modifications to the CoMi pattern on specific biological processes are a useful representation of drug action by predicting the response to neoadjuvant Paclitaxel treatment in breast cancer and show that the drug counteracts the CoMi network dysregulation induced by tumorigenesis. We then generate a quantitative testbed to investigate the ability of the CoMi pattern to distinguish FDA approved breast cancer drugs from other FDA approved drugs not related to breast cancer. We also compare the ability of the CoMi and metagene methods to predict response to neoadjuvant Paclitaxel treatment in clinical cohorts. We find the CoMi method outperforms the metagene method, achieving area under curve (AUC) values of 0.78 and 0.66 respectively. Furthermore, several of the predicted CoMi features highlight the network-based mechanism of drug resistance. Thus, our study suggests that explicitly modeling the drug action using network biology provides a promising approach for predictive marker discovery.

Environmental and genetic factors that contribute to Escherichia coli K-12 biofilm formation

Biofilms are communities of bacteria whose formation on surfaces requires a large portion of the bacteria's transcriptional network. To identify environmental conditions and transcriptional regulators that contribute to sensing these conditions, we used a high-throughput approach to monitor biofilm biomass produced by an isogenic set of Escherichia coli K-12 strains grown under combinations of environmental conditions. Of the environmental combinations, growth in tryptic soy broth at 37 degrees C supported the most biofilm production. To analyze the complex relationships between the diverse cell-surface organelles, transcriptional regulators, and metabolic enzymes represented by the tested mutant set, we used a novel vector-item pattern-mining algorithm. The algorithm related biofilm amounts to the functional annotations of each mutated protein. The pattern with the best statistical significance was the gene ontology 'pyruvate catabolic process,' which is associated with enzymes of acetate metabolism. Phenotype microarray experiments illustrated that carbon sources that are metabolized to acetyl-coenzyme A, acetyl phosphate, and acetate are particularly supportive of biofilm formation. Scanning electron microscopy revealed structural differences between mutants that lack acetate metabolism enzymes and their parent and confirmed the quantitative differences. We conclude that acetate metabolism functions as a metabolic sensor, transmitting changes in environmental conditions to biofilm biomass and structure.

Relating gene expression data on two-component systems to functional annotations in Escherichia coli

Background

Obtaining physiological insights from microarray experiments requires computational techniques that relate gene expression data to functional information. Traditionally, this has been done in two consecutive steps. The first step identifies important genes through clustering or statistical techniques, while the second step assigns biological functions to the identified groups. Recently, techniques have been developed that identify such relationships in a single step.

Results

We have developed an algorithm that relates patterns of gene expression in a set of microarray experiments to functional groups in one step. Our only assumption is that patterns co-occur frequently. The effectiveness of the algorithm is demonstrated as part of a study of regulation by two-component systems in Escherichia coli. The significance of the relationships between expression data and functional annotations is evaluated based on density histograms that are constructed using product similarity among expression vectors. We present a biological analysis of three of the resulting functional groups of proteins, develop hypotheses for further biological studies, and test one of these hypotheses experimentally. A comparison with other algorithms and a different data set is presented.

Conclusion

Our new algorithm is able to find interesting and biologically meaningful relationships, not found by other algorithms, in previously analyzed data sets. Scaling of the algorithm to large data sets can be achieved based on a theoretical model.