Adam Officer, a graduate student in Olivier Harismendy’s and Pablo Tamayo’s lab (UC San Diego). Ductal carcinoma in situ (DCIS) is a frequently detected pre-cancerous lesion in breast tissue. As there are no reliable markers for risk of progression to breast cancer, it is likely that most DCIS are overtreated. Previous studies have found little to no differences in mRNA expression levels in breast epithelium from DCIS and invasive ductal carcinoma (IDC), suggesting that changes in cell signaling in the surrounding environment play a key role in progression. In this project, I will perform an unbiased analysis of microenvironment changes in the transition from DCIS to IDC. Using network propagation and functional annotations of ligand-receptor interactions, I hope to improve methods for identifying differential cell-cell interactions, potentially revealing mechanisms of progression and biomarkers.
Alex Tankka, a graduate student in Gene Yeo’s lab (UC San Diego). The c-MYC (MYC) transcription factor is the most frequently amplified oncogene in human cancers. Direct pharmacological targeting of MYC is challenging though because it regulates key processes in both normal and tumor cells. An alternative strategy is to identify gene functions that are essential only for MYC-driven tumor cells. There is increasing evidence that MYC regulates the expression of microRNAs (miRs), an abundant class of small non-protein-coding RNAs that serve as regulators of diverse cohorts of target genes in tissue and cell context dependent manners. This project is based on the hypothesis that uncharacterized miR-mRNA interactions are required for MYC-dependent cancer cells to maintain their oncogenic gene expression programs. Since individual proteins rarely mediate cellular behaviors, this project will employ a systems biology approach to reveal the miR-mediated gene silencing pathways using large scale-functional studies.
Alex Wenzel, a graduate student in Jill Mesirov’s lab (UC San Diego). Deriving knowledge of pathway activity from mRNA expression data requires methods such as Gene Set Enrichment Analysis (GSEA). Some gene sets lack sensitivity and specificity for reasons such as the subjectivity of manual curation or subsets of genes within the set that are both significantly up- and down-regulated. I recently developed a data-driven approach to refining existing gene sets, and although I found improvements in most cases, some gene sets became too small for practical use. I hypothesize that there may be other genes that if included with the initial gene set would better represent the targeted phenotype or cellular state after refinement. In this project, I will use a network diffusion approach with curated protein-protein interaction networks to identify neighboring genes to include prior to gene set refinement, with the goal of increasing the sensitivity, specificity and completeness of the refined gene sets.
Antoine Forget, an Associate Researcher in Nevan Krogan’s lab (UCSF). Phosphoinositide 3-kinases (PI3Ks) are central regulators of cell proliferation, growth and survival, and are among the most recurrently mutated genes in cancer. Although over twenty PI3K-inhibitors are in clinical development for treating cancers, broad-specificity PI3K inhibitors have strong negative metabolic side effects due to the essential role of PI3Ks in signaling. Therefore, it is critical both to consider the cell type and mutational context of PI3Ks and to identify context-specific combinatorial therapies synergistic with PI3Ks inhibition to obtain durable responses. A pan-cancer study of PI3K’s cell specific regulation and dependencies will provide critical insights into the design of therapeutic strategies by tackling toxicity and resistance. Here, I propose to map the functional network and genetic co-dependencies of PIK3CA, the most mutated member of the PI3K family in cancer, to uncover cancer-specific vulnerabilities associated with the deregulation of the PI3K pathway.
Mark Magbanua, a Senior Scientist in Laura van ‘t Veer’s lab (UCSF). The presence of circulating tumor DNA (ctDNA) in the blood of untreated cancer patients is associated with aggressive disease and portends poor clinical outcomes. The underlying biological mechanisms involved in ctDNA shedding and clearance during therapy are poorly understood. I hypothesize that molecular markers predictive of the presence of ctDNA in blood, identified from modeling of tumor molecular data, can illuminate mechanisms involved in ctDNA shedding and clearance during treatment. To address this hypothesis, I will use machine learning approaches to discover genomic predictors of ctDNA shedding in early breast cancer patients treated with neoadjuvant chemotherapy (NAC). In a similar manner, I will characterize genomic predictors of ctDNA kinetics early during the course of treatment. I will then use a systems biology approach to uncover gene networks and pathways associated with ctDNA shedding and clearance. The overall goal is to shed light on the biology of ctDNA shedding and kinetics during therapy in the context of tumor heterogeneity and response to treatment.
Yue Qin, a graduate student in Trey Ideker’s lab (UC San Diego). Cancer research is in the midst of a data revolution, with numerous studies having generated large-scale data on the tumor transcriptome, proteome, metabolome, imaging and so on. However, it is yet unclear how to gain a holistic view of cancer cells by effectively combining data that capture different information at different scales. To address this challenge, I recently developed a method to create a multi-scale cell map by integrating immunofluorescence images and affinity purification-mass spectrometry data. In my current work, I aim to develop a visible neural network (VNN) using this map to investigate the molecular mechanisms connecting genotypes to phenotypes. I hypothesize that by predicting cancer cell fitnesses given a genomic context and a single gene deletion, a VNN can learn to predict pairwise synthetic lethal interactions, which can be used for targeted treatment decisions. The incorporation of genomic context in my VNN model design enables generalizability of this model to various types of cancer.