Dataset Information

Harnessing the biological complexity of Big Data from LINCS gene expression signatures.

ABSTRACT: Gene expression profiling using transcriptional drug perturbations are useful for many biomedical discovery studies including drug repurposing and elucidation of drug mechanisms (MoA) and many other pharmacogenomic applications. However, limited data availability across cell types has severely hindered our capacity to progress in these areas. To fill this gap, recently, the LINCS program generated almost 1.3 million profiles for over 40,000 drug and genetic perturbations for over 70 different human cell types, including meta information about the experimental conditions and cell lines. Unfortunately, Big Data like the ones generated from the ongoing LINCS program do not enable easy insights from the data but possess considerable challenges toward their analysis. In this paper, we address some of these challenges. Specifically, first, we study the gene expression signature profiles from all cell lines and their perturbagents in order to obtain insights in the distributional characteristics of available conditions. Second, we investigate the differential expression of genes for all cell lines obtaining an understanding of condition dependent differential expression manifesting the biological complexity of perturbagents. As a result, our analysis helps the experimental design of follow-up studies, e.g., by selecting appropriate cell lines.

SUBMITTER: Musa A

PROVIDER: S-EPMC6114505 | biostudies-literature | 2018

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Harnessing the biological complexity of Big Data from LINCS gene expression signatures.

Musa Aliyu A Tripathi Shailesh S Kandhavelu Meenakshisundaram M Dehmer Matthias M Emmert-Streib Frank F

PloS one 20180829 8

Gene expression profiling using transcriptional drug perturbations are useful for many biomedical discovery studies including drug repurposing and elucidation of drug mechanisms (MoA) and many other pharmacogenomic applications. However, limited data availability across cell types has severely hindered our capacity to progress in these areas. To fill this gap, recently, the LINCS program generated almost 1.3 million profiles for over 40,000 drug and genetic perturbations for over 70 different hu ...[more]

PMID: 30157183

Similar Datasets

Project description:BackgroundLiver hepatocellular carcinoma accounts for the overwhelming majority of primary liver cancers and its belated diagnosis and poor prognosis call for novel biomarkers to be discovered, which, in the era of big data, innovative bioinformatics and computational techniques can prove to be highly helpful in.MethodsBig data aggregated from The Cancer Genome Atlas and Natural Language Processing were integrated to generate differentially expressed genes. Relevant signaling pathways of differentially expressed genes went through Gene Ontology enrichment analysis, Kyoto Encyclopedia of Genes and Genomes and Panther pathway enrichment analysis and protein-protein interaction network. The pathway ranked high in the enrichment analysis was further investigated, and selected genes with top priority were evaluated and assessed in terms of their diagnostic and prognostic values.ResultsA list of 389 genes was generated by overlapping genes from The Cancer Genome Atlas and Natural Language Processing. Three pathways demonstrated top priorities, and the one with specific associations with cancers, 'pathways in cancer,' was analyzed with its four highlighted genes, namely, BIRC5, E2F1, CCNE1, and CDKN2A, which were validated using Oncomine. The detection pool composed of the four genes presented satisfactory diagnostic power with an outstanding integrated AUC of 0.990 (95% CI [0.982-0.998], P < 0.001, sensitivity: 96.0%, specificity: 96.5%). BIRC5 (P = 0.021) and CCNE1 (P = 0.027) were associated with poor prognosis, while CDKN2A (P = 0.066) and E2F1 (P = 0.088) demonstrated no statistically significant differences.DiscussionThe study illustrates liver hepatocellular carcinoma gene signatures, related pathways and networks from the perspective of big data, featuring the cancer-specific pathway with priority, 'pathways in cancer.' The detection pool of the four highlighted genes, namely BIRC5, E2F1, CCNE1 and CDKN2A, should be further investigated given its high evidence level of diagnosis, whereas the prognostic powers of BIRC5 and CCNE1 are equally attractive and worthy of attention.

Dataset Information

Harnessing the biological complexity of Big Data from LINCS gene expression signatures.

Publications

Harnessing the biological complexity of Big Data from LINCS gene expression signatures.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets