Cell-Type-Specific Data Integration and Machine-Learning Methodology.

<p>Our approach is designed to generate reliable and relevant predictive biological networks using high-throughput data limited to a specific cell type and a training gold standard focused on biological processes active in that cell type. This process can be distilled into four basic steps: <b>1.</b> Collect and standardize cell-type specific data from studies using diverse high-throughput experimental techniques, including microarray gene expression, chromatin immunoprecipitation (ChIP) on chip (ChIP-Chip), ChIP followed by high-throughput-sequencing (ChIP-Seq), affinity purification followed by mass spectrometry (AP-MS), whole-genome small interfering RNA (siRNA) screens, and phylogenetic sequence similarity. For this case study, we focused on mouse embryonic stem cell (mESC) data. <b>2.</b> Curate a process-specific gold standard training set to provide a baseline for assessing data reliability and significance for related biological processes known to be active in the cell type of interest. Our gold standard training set consists of experimentally validated pair-wise associations between genes and proteins known to be involved in mESC self-renewal, pluripotency, and cell fate determination. <b>3.</b> Iteratively test and validate networks. <i>A.</i> Use a naïve Bayesian network classifier to perform inference and predict novel gene and protein relationships. Our network predicts pairwise functional associations that influence mESC self-renewal and early developmental processes. <i>B.</i> Validate the accuracy of predicted functional relationships using standard machine learning performance metrics, cross validation, and bootstrapping, followed by evaluation of biological content. Our protocol for assessing networks ensures our results are highly reliable and relevant to mESC self-renewal. <b>4.</b> Provide community access to analyses and tools. Through StemSight.org, we provide access to network analyses and visualization tools to enable users to further explore networks centered on their genes of interest.