Pathway selection criteria and the overall pathway sets collected in this study are listed in Additional file 1. Our goal was to use protein interactions and regula tory reactions assembled into metabolic pathways with out introducing duplicated links and elements. To merge interactions from various sources, the genes alias names must be arranged in advance. Furthermore, we recorded the directions of interactions between genes as well to the graph. We joined the pro teins as vertices to the integrated large network and connected them to any co regulated genes by adding new edges. From a biological viewpoint of transcrip tional relationships, a number of genes may regulate themselves or regulate each other, resulting in cyclic relationships while re constructing the large network, which makes it more difficult to determine simple short est paths.

We dealt with this problem by merging ver tices as demonstrated in Figure 1. Taking Figure 1 as an example, the transcription factors AR and DDIT3 regulate their target genes and regulate each other as well. To preserve the biological truth and avoid loops being represented in the graph, vertices AR and DDIT3 were merged during the shortest paths algorithm. Next, while scoring the identified pathways according to gene expression data, each vertex was considered separately and identically. Microarray data Peters et al. presented the results of a preliminary inves tigation into the molecular phenotype of patient derived ovarian tumor cells in the context of sensitivity or resis tance to carboplatin.

They correlated chemore sponse data with gene expression patterns at the level of transcription. Primary cultures of cells derived GSK-3 from ovarian carcinomas of individual patients were characterized using the ChemoFx assay and classified as either carboplatin sensitive or resistant. Three representative cultures of cells from each indivi dual tumor were then subjected to Affymetrix gene chip analysis using U95A human gene chip arrays. They identified numbers of differentially expressed genes that define transcriptional differences between chemosensitive and chemoresistant cells and temporal responses to carboplatin expressed in an ex vivo setting. Gabriela et al. investigated the response to cisplatin of a panel of NSCLC cell lines and found an inverse correla tion between sensitivity and damage formation resulting from this agent.

Further analysis of multiple alter nate cellular end points including cell cycle analysis, apoptosis and gene expression changes, revealed cispla tin damage tolerance to be a mechanism of chemoresis tance in this model system. Both gene expression data sets were available through the Gene Expression Omni bus at NCBI. Systems and implementation System overview A system flow diagram of the corresponding processes is shown in Figure 2.