All viruses rely on cellular signaling pathways and metabolic programs to make copies of themselves. Classic examples are ribosomal translation, ATP generation, and nucleotide biosynthesis. All cells translate proteins and generate ATP and most can either import or sythesize nucleotides. Yet, viruses cannot infect all cells. Each virus has a defined "tropism", or capacity to infect different cell types.

Numerous examples of non-permissive and semi-permissive infections have been observed in experimental infection systems, illustrating that some cell types posses intrinsic barriers to infection. In many cases, non-permissive cell types simply lack surface entry factors required to internalize virus particles, but in others virus particles successfully enter cells, but the infectious cycle stalls. The existence of these "semi-permissive" infections implies that cellular pathways besides translation and ATP generation are required to support viral replication; and that limiting cellular functions downstream of entry can be identified.

We are pursuing two general approaches to identify new cellular pathways required for virus replication: (1) computational analysis of gene-expression patterns in virally-infected cells, and (2) restoration of essential "host-factors" in non-permissive and semi-permissive cell types.

Gene-expression profiling is a powerful methodology for determining how cells respond to viral infection. Recent technologies have enabled routine acquisition of global gene-expression patterns (transcriptomes) and sophisticated computational methods have been developed to monitor how sets of genes coordinately respond to different experimental treatments. Databases of thousands of gene-expression experiments, across many biological domains, have accumulated and are accessible to biomedical researchers. This wealth of data can potentially be used to extract useful biological information through cross-study re-analysis, or "meta-analysis."

However, public data is challenging to use for discovery. An apical bottleneck to data reuse, is metadata organization and standardization. Public databases, such as the NCBI's Gene-Expression Omnibus (GEO), require submission of metadata describing each sample's biological parameters (e.g. "cell-type", "treatment", or "replicate"). However, due to the vastly different types of experiments submitted to such databases, rigid metadata vocabularies have not been established. Therefore, it can be difficult to organize many different studies and metadata typically needs to be coerced into a consistent format, on an ad-hoc basis, for re-analysis. This is especially important when analyzing dozens to hundreds of data sets or scripting automated processing routines.

We are developing several tools to aid in acquisition, organization, and standardization of sequencing metadata. These tools enable metadata organization for single to hundreds of sequencing studies, can be deployed by individual users or teams of researchers, and allow preparation of customized metadata databases that can be easily parsed by automated data processing routines. We intend to use these tools to perform large meta-analyses of both public and newly acquired gene-expression data sets.