Estella Yixing DONG

• Integrating scRNAseq and spatial transcriptomics (ST) to decode breast, lung, DLBCL cancer.
• Leveraging deep learning-extracted imaging features to inform gene expression analysis.
• Developing pipeline for various ST technologies, such as GeoMx, Visium, VisiumHD & Xenium.

• Led first R for submission at Bayer with pharmaverse admiral, managing scope of 2 sub-teams.
• Designed templates that streamlined AE reporting in CSR; drafted developer’s guidelines.
• Assist open-source initiatives in the department and industry-wide working groups.

• Used IPTW to create external control cohorts for a Phase II oncology trial.
• Predicted the remission of rare diseases with random survival forest.
• Developed R pipelines to automate chart review quality reports, improved efficiency.
• Analyzed autoimmune disorder characteristics using IBM Optum NLP data.

• Built Shiny module to explore AE and efficacy of a Phase I FDA breakthrough molecule.
• Audited Phase II/III trial meetings, interim analysis, and reviewed study protocol and SAP.
• Developed R functions for disease biology analysis of RNA-seq data from 8 NHL trials.
• Created network analysis animation to identify real-world drug use patterns.

• Pre-processed H&E and Feulgen images in MATLAB; tuned CNNs in Python.
• Trained Random Forest classifiers in R on imbalanced oral cell data to assist pathologists.
• Delivered bi-monthly presentations on deep learning applications in pathology.

• Evaluated abstract submissions (Biostatistics, Machine Learning) from conference participants.

• Co-authored a white paper on OSS license guidance for the pharmaceutical industry.

• Supported student activities and documented committee meeting minutes.
• Proposed and invited speakers to Canadian Statistics Student Conference workshop session.

• Organized events, including career nights, programming workshops, and student meetups.
• Guided two junior students in academic and curriculum planning.

• Integrated causal inference & survival analysis for longitudinal pharmacoepidemiology.
• Used weighted case-base sampling as an alternative to Cox proportional hazards modeling.

• Trained CNNs with transfer learning and compared to LDA for classifying ErrP EEG signals.

• Simulated ground truth for Steven power law in fMRI BOLD signals via a Bayesian framework.

• Awarded to academic speakers in the conferences.

• Nominated by colleagues for project contributions and leadership, with a 2nd-highest award.

• Awarded to students to participate in R in Pharma 2019 conference in Boston.

• Awarded as one of the 25 funded scholars among 600 applicants globally.

• Awarded to undergraduate students who participate in Research Abroad.

• Awarded with one 1st & two 2nd-highest honors, totaling $25,000.