The increasing availability of electronic health record (EHR) systems has created enormous potential for translational research. However, it is difficult to know all the relevant codes related to a phenotype due to the large number of codes available. Traditional data mining approaches often require the use of patient-level data, which hinders the ability to share data across institutions. In this project, we demonstrate that multi-center large-scale code embeddings can be used to efficiently identify relevant features related to a disease of interest. We constructed large-scale code embeddings for a wide range of codified concepts from EHRs from two large medical centers. We developed knowledge extraction via sparse embedding regression (KESER) for feature selection and integrative network analysis. We evaluated the quality of the code embeddings and assessed the performance of KESER in feature selection for eight diseases. Besides, we developed an integrated clinical knowledge map combining embedding data from both institutions. The features selected by KESER were comprehensive compared to lists of codified data generated by domain experts. Features identified via KESER resulted in comparable performance to those built upon features selected manually or with patient-level data. The knowledge map created using an integrative analysis identified disease-disease and disease-drug pairs more accurately compared to those identified using single institution data. Analysis of code embeddings via KESER can effectively reveal clinical knowledge and infer relatedness among codified concepts. KESER bypasses the need for patient-level data in individual analyses providing a significant advance in enabling multi-center studies using EHR data.

Purpose: Paclitaxel (PTX) is one the most potent and commonly used chemotherapies for breast and pancreatic cancer. Several ongoing clinical trials are investigating means of enhancing delivery of PTX across the blood-brain barrier for glioblastomas (GBMs). Despite the widespread use of PTX for breast cancer, and the initiative to repurpose this drug for gliomas, there are no predictive biomarkers to inform which patients will likely benefit from this therapy. Experimental Design: To identify predictive biomarkers for susceptibility to PTX, we performed a genome-wide CRISPR knock-out (KO) screen using human glioma cells. The genes whose KO was most enriched in the CRISPR screen underwent further selection based on their correlation with survival in the breast cancer patient cohorts treated with PTX and not in patients treated with other chemotherapies, a finding that was validated on a second independent patient cohort using progression-free survival. Results: Combination of CRISPR screen results with outcomes from taxane-treated breast cancer patients led to the discovery of endoplasmic reticulum (ER) protein SSR3 as a putative predictive biomarker for PTX. SSR3 protein levels showed positive correlation with susceptibility to PTX in breast cancer cells, glioma cells and in multiple intracranial glioma xenografts models. Knockout of SSR3 turned the cells resistant to PTX while its overexpression sensitized the cells to PTX. Mechanistically, SSR3 confers susceptibility to PTX through regulation of phosphorylation of ER stress sensor IRE1α. Conclusion: Our hypothesis generating study showed SSR3 as a putative biomarker for susceptibility to PTX, warranting its prospective clinical validation.

No abstract

Paclitaxel (PTX) is one the most potent and commonly used chemotherapies for breast and pancreatic cancer. Given the potency of this drug for glioblastomas (GBM) several ongoing clinical trials are investigating means of enhancing delivery of PTX across the blood-brain barrier for this disease. In spite of the efficacy of PTX, individual tumors exhibit variable susceptibility to this drug, with response rate in the range of 30%-60%. To identify predictive biomarkers for response to PTX, we performed a genome-wide CRISPR knock-out screen using human glioma cells. The most enriched genes in the CRISPR screen underwent further selection based on their correlation with survival in the breast cancer patient cohorts treated with PTX and not in patients treated with other chemotherapies, a finding that was validated on a second independent patient cohort. This led to the discovery of endoplasmic reticulum (ER) protein SSR3 as a putative predictive biomarker for PTX. SSR3 protein levels showed positive correlation with response to PTX in breast cancer cells, glioma cells, in multiple intracranial glioma xenografts and in GBM patient derived explant cultures. Knockout of SSR3 turned the cells resistant to PTX while its overexpression sensitized the cells to PTX. In gliomas, SSR3-mediated susceptibility to PTX relates to modulation of phosphorylation of ER stress sensor IRE1α. Thus, by using genome-wide screen combined with patient response data, we discovered a biomarker that demonstrates causal and correlative relationship with response to PTX in breast cancer and GBM. Prospective validation of this biomarker is warranted for its broad implementation for precision oncology.