Date of Award
5-10-2026
Date Published
June 2026
Degree Type
Thesis
Degree Name
Master of Science (MS)
Department
Biomedical and Chemical Engineering
Advisor(s)
Zhen Ma
Keywords
AI;Bioengineering;Cardiomyocyte;Deep Learning;Machine Learning;Stem Cell
Abstract
Recent advances in human induced pluripotent stem cell (hiPSC) technology have enabled the development of high-throughput platforms for cardiotoxicity screening. However, conventional analytical approaches remain limited by time-intensive processing, manual annotation requirements, and reduced scalability for large datasets. To address these challenges, calcium fluorescence signals from human induced pluripotent stem cell-derived cardiomyocytes (hiPSC CMs) were analyzed using a high-throughput imaging platform and complementary machine learning frameworks. First, a feature-based supervised learning approach using engineered waveform descriptors and low-dimensional signal embeddings enabled rapid and interpretable screening of large chemical datasets, with combined features improving predictive performance relative to engineered features alone. Second, an unsupervised autoencoder-based multi-donor voting platform was developed to quantify drug-induced cardiotoxicity directly from raw waveforms without manual labeling. Application of this framework revealed donor-specific variability in drug responses across multiple hiPSC lines, while aggregation of donor predictions enabled population-level IC50 estimation, safety margin analysis, and consensus toxicity assessment. Together, these findings demonstrate a scalable computational strategy for cardiotoxicity screening that integrates rapid classification, waveform-level toxicity detection, and biologically meaningful donor variability.
Access
Open Access
Recommended Citation
Vu, Danny, "Data-Driven Approaches to Investigate Cardiotoxicity using hiPSCs High Throughput Screening" (2026). Theses - ALL. 1006.
https://surface.syr.edu/thesis/1006
