Infectious diseases are the second leading cause of death worldwide, next to cardiovascular diseases 
. Bacterial infections such as tuberculosis and food- and water- borne infections from Salmonella enterica
and Escherichia coli
still present many challenges to biomedical researchers. Foremost among these challenges is that infectious agents rapidly mutate and become resistant to drugs 
. The conventional approach of targeting pathogen proteins has accelerated the spread of resistance, resulting in the re-emergence of once-contained infectious diseases, such as those caused by multidrug-resistant strains of Mycobacterium tuberculosis, Staphylococcus aureus
, and Salmonella enterica
. In an effort to combat the issue of drug resistance, anti-infective drug discovery is shifting to a new approach that targets the host instead of pathogens 
. “Host-oriented” drug discovery focuses on manipulating or subverting biological processes in the host that pathogens utilize 
. Another problem facing the treatment of infectious diseases is the increasing number of pathogenic agents 
. Furthermore, new pathogens are appearing regularly, e.g., the pandemic swine flu H1N1 virus recognized in 2009. The expanding range of infectious agents coupled with the high cost associated with drug discovery have made it economically infeasible and practically impossible to tackle each pathogen individually 
. These factors have necessitated treatment regimens that are effective against a wide variety of infectious agents.
These factors have encouraged efforts in host-oriented broad-spectrum (HOBS) drug discovery, i.e., finding targets in the host that can simultaneously cure multiple infections 
. Examples of HOBS drugs currently available in the market include Statins and Isoprinosine. Statins are used in the treatment of Leishmania, Staphylococcus aureus
, and HIV infections 
. Statins lower the cholesterol level in human body. They are effective against pathogens that utilize cholesterol in binding and internalization to the host cell. Isoprinosine, which stimulates the proliferation of T-cells, is used in the treatment of Herpes simplex
, and Epstein-Barr
virus infections 
A first and important step in HOBS drug discovery is the development of computational tools to discover common physiological processes and cellular pathways that different pathogens utilize to infect, proliferate, and spread in the host. We hypothesized that comprehensive molecular datasets of host responses to diverse varieties of pathogens might form a powerful resource to discover such pathways. Transcriptional datasets that correspond to different infectious diseases, cell/tissue types, and organisms are the most abundantly available. Meta-analysis of transcriptional datasets have been performed for a wide range of diseases. For instance, Rhodes et al.
analyzed 40 cancer related microarray datasets to identify common signatures of cancer. English and Butte 
integrated 49 obesity-related genome-wide experiments obtained from human, mouse, rat, and worm to predict new genes that may be associated with obesity. Magalhaes et al.
performed meta-analysis of 27 age-related gene expression profile datasets from human, mouse, and rat to reveal several common signatures of aging. Jenner et al.
used hierarchical clustering of gene expression profiles of 77 pathogens in order to find genes that exhibited similar expression profiles across several disease types.
Recent approaches have taken meta-analysis of DNA microarray datasets one step further by incorporating drug targets into the analysis and inferring new uses for existing drugs on the basis of disease similarities. The premise underlying these approaches is that diseases with a high degree of transcriptional similarity might be treated with similar drugs 
. Hu et al.
discovered disease-disease links by using correlation-based methods and gene set enrichment analysis to measure the similarities between gene expression profiles of diseases. They also integrated gene expression profiles that pertain to responses of cell lines to drugs derived from the Connectivity Map 
to create a drug-disease network where clusters of drugs and diseases suggested shared drug mechanisms and molecular disease pathology. Suthram et al.
performed integrative analysis of 54 disease-related mRNA expression datasets. They measured the perturbation of predefined protein functional modules using the mean normalized transcriptional activity of each module's component genes in the disease's transcriptional profile. Furthermore, they identified known drug targets in the modules that were perturbed by multiple disease types, which they proposed as pluripotent/broad-spectrum drug targets .
The goal of our work is similar to that of Jenner et al. , Hu et al. , and Suthram et al. : to discover transcriptional responses common to many diseases, specifically those caused by bacterial pathogens, and to discover existing drug targets within those transcriptional signatures. The previous authors have used global correlation measures to detect disease associations, which may obscure relationships that exist over only a subset of the diseases or genes. In contrast, we use a combination of gene set level enrichment and biclustering. As we demonstrate in this work, this approach enables us to group sets of host genes that are dysregulated only by a subset of the pathogens, facilitating the capture of pathway-specific relationships among groups of pathogens.