BACKGROUND: The course of disease for patients with idiopathic pulmonary fibrosis (IPF) is highly heterogeneous. Prognostic models rely on demographic and clinical characteristics and are not reproducible. Integrating data from genomic analyses may identify novel prognostic models and provide mechanistic insights into IPF. METHODS: Total RNA of peripheral blood mononuclear cells was subjected to microarray profiling in a training (45 IPF individuals) and two independent validation cohorts (21 IPF/10 controls, and 75 IPF individuals, respectively). To identify a gene set predictive of IPF prognosis, we incorporated genomic, clinical, and outcome data from the training cohort. Predictor genes were selected if all the following criteria were met: 1) Present in a gene co-expression module from Weighted Gene Co-expression Network Analysis (WGCNA) that correlated with pulmonary function (p < 0.05); 2) Differentially expressed between observed "good" vs. "poor" prognosis with fold change (FC) >1.5 and false discovery rate (FDR) < 2 %; and 3) Predictive of mortality (p < 0.05) in univariate Cox regression analysis. "Survival risk group prediction" was adopted to construct a functional genomic model that used the IPF prognostic predictor gene set to derive a prognostic index (PI) for each patient into either high or low risk for survival outcomes. Prediction accuracy was assessed with a repeated 10-fold cross-validation algorithm and independently assessed in two validation cohorts through multivariate Cox regression survival analysis. RESULTS: A set of 118 IPF prognostic predictor genes was used to derive the functional genomic model and PI. In the training cohort, high-risk IPF patients predicted by PI had significantly shorter survival compared to those labeled as low-risk patients (log rank p < 0.001). The prediction accuracy was further validated in two independent cohorts (log rank p < 0.001 and 0.002). Functional pathway analysis revealed that the canonical pathways enriched with the IPF prognostic predictor gene set were involved in T-cell biology, including iCOS, T-cell receptor, and CD28 signaling. CONCLUSIONS: Using supervised and unsupervised analyses, we identified a set of IPF prognostic predictor genes and derived a functional genomic model that predicted high and low-risk IPF patients with high accuracy. This genomic model may complement current prognostic tools to deliver more personalized care for IPF patients.

Gene expression profiles are some of the most abundant sources of data about the cellular state of a collection of cells in an organism. Comparison of the expression profiles of multiple samples allows biologists to find associations between observations at the molecular level and the phenotype of the samples. A key challenge is to distinguish variation in expression due to biological factors of interest from variation due to confounding factors that can arise for unrelated technical or biological reasons. This thesis presents models that can explicitly adjust the comparison of expression profiles to account for specific types of confounding factors. One such confounding factor arises when comparing tissue-specific expression profiles across multiple organisms to identify differences in expression that are indicative of changes in gene function. When the organisms are separated by long evolutionary distances, tissue functions may be re-distributed and introduce expression changes unrelated to changes in gene function. We developed Brownian Factor Phylogenetic Analysis, a model that can account for such re-distribution of function, and demonstrate that removing this confounding factor improves tasks such as predicting gene function. Another confounding factor arises because current protocols for expression profiling require RNA extracts from multiple cells. Often biological samples are heterogeneous mixtures of multiple cell types, so the measured expression profile is an average of the RNA levels of the constituent cells. When the biological sample contains both cells of interest and nuisance cells, the confounding expression from the nuisance cells can mask the expression of the cells of interest. We developed ISOLATE and ISOpure, two models for addressing the heterogeneity of tumor samples. We demonstrated that modeling tumor heterogeneity leads to an improvement in two tasks: identifying the site of origin of metastatic tumors, and predicting the risk of death of lung cancer patients.

Gene expression profiles are some of the most abundant sources of data about the cellular state of a collection of cells in an organism. Comparison of the expression profiles of multiple samples allows biologists to find associations between observations at the molecular level and the phenotype of the samples. A key challenge is to distinguish variation in expression due to biological factors of interest from variation due to confounding factors that can arise for unrelated technical or biological reasons. This thesis presents models that can explicitly adjust the comparison of expression profiles to account for specific types of confounding factors. One such confounding factor arises when comparing tissue-specific expression profiles across multiple organisms to identify differences in expression that are indicative of changes in gene function. When the organisms are separated by long evolutionary distances, tissue functions may be re-distributed and introduce expression changes unrelated to changes in gene function. We developed Brownian Factor Phylogenetic Analysis, a model that can account for such re-distribution of function, and demonstrate that removing this confounding factor improves tasks such as predicting gene function. Another confounding factor arises because current protocols for expression profiling require RNA extracts from multiple cells. Often biological samples are heterogeneous mixtures of multiple cell types, so the measured expression profile is an average of the RNA levels of the constituent cells. When the biological sample contains both cells of interest and nuisance cells, the confounding expression from the nuisance cells can mask the expression of the cells of interest. We developed ISOLATE and ISOpure, two models for addressing the heterogeneity of tumor samples. We demonstrated that modeling tumor heterogeneity leads to an improvement in two tasks: identifying the site of origin of metastatic tumors, and predicting the risk of death of lung cancer patients.