Search

The detection and quantification of fusion transcripts has both biological and clinical implications. RNA sequencing provides a means for unbiased and high resolution characterization of fusion transcript information in tissue samples. We evaluated two gene fusion detection algorithms, TopHat-Fusion and deFuse, using transcriptomes derived from breast cancer biopsies and cell line. The two algorithms use different strategies to reduce false positive fusion calls: TopHat-Fusion uses a cascade of heuristically designed filters while deFuse relies on statistical methodologies to assess and prune potential fusions. We evaluated the concordance of gene fusion calls between the two algorithms and characterized the evidence generated by each methodology for the common fusions. Our results show that the concordance between the two algorithms is about 6% at comparable number of fusions detected. deFuse tends to be more sensitive in detecting fusions when compared to TopHat-Fusion. On average, deFuse identifies proportionally more reads than TopHat-Fusion as evidence for concordant fusions. While deFuse provides a more quantitative summary of the reliability of the fusion calls, TopHat-Fusion is able to detect additional fusion transcripts that were missed by deFuse. Further research is required to develop algorithms that can detect fusions with high sensitivity and specificity.

Whole genome sequencing enables a high resolution view of the human genome and enables unique insights into copy number variations on an unprecedented scale. Numerous tools and studies have already been introduced that provide confirmatory evidence and new genomic structure variation data in individuals as well as across populations. We utilize two such tools, CNV-seq and FREEC to compare their outputs when applied to five whole genome sequences representing four populations. We focus on the ability of these tools to detect segments from two sets of segments known to vary across populations, and discuss the direction and the challenges in developing tools that detect copy number variation in collections of human genomes.

Ovarian cancer is the leading cause of death in gynecological cancers. Carboplatinum-based therapy is the standard treatment choice for ovarian cancer. However, a majority of the patients develop resistance to carboplatinum fairly rapidly hence there is a clinical need for early predictors of carboplatinum resistance. While there are a few indicative gene markers, they have poor sensitivity and specificity in predicting response accurately. It is essential that multiple high throughput molecular profiling modalities are integrated and investigated to provide a full picture of the ongoing processes. Here, we propose a methodology to identify central players in platinum resistance from a list of stratifying genes using a data-driven approach. We have used correlation of DNA methylation and gene expression data and applied network based features to identify the influence of DNA methylation on gene expression. This provides interpretive analysis and is complementary to the biological pathway-enrichment approaches. We suggest that our method, based on network structure properties, adds a useful layer to multi-modal evidence integration to focus on the key processes and interactions in resistance mechanisms.

Genomic signatures for disease prognosis are discovered by applying statistical methods on high-throughput data. Enrichment analysis of gene set annotations provide limited understanding of the biological processes. We use curated sets reflecting various biological processes implicated in cancer and the thematic clusters to describe and compare four breast cancer prognostic signatures. We demonstrate an effective and automated use of curated gene sets to identify thematic overlaps in gene sets based on their associated significant Gene Ontology terms. This method enabled us to compare four breast cancer prognostic signatures The differences in the thematic content suggest that the signatures while answering the issue of aggressiveness are capturing different sets of biological processes to achieve efficacy.

It is widely acknowledged in the bioinformatics community that many of the available datasets used for discovering possible molecular diagnostic patterns have very low case-to-feature ratios and this leads to problems of high false positive discovery rates. We (and others) contend that these ratios are so low that the usual methods of cross-validation re-sampling etc. will never eliminate the problem. We conjecture that additional knowledge sources are needed to augment the statistical pattern discovery. We describe our initial attempts to do this by combining information from the Gene Ontology with multiple sets of discriminating patterns discovered in a publicly available gene expression dataset on breast cancer. Using a second validation dataset we identify some "good" and "poor" validators among the candidate patterns and then show that certain biological contexts appear to provide some ability to discriminate between them. We also describe some approaches we took to dealing with the non-uniformity in online biological data banks and the difficulties this creates when seeking information on a large number of genes. While we do not claim to have proven the value of this approach we are encouraged by these initial findings.