Text Mining Applications in Leukemia
Loading...
Author(s)
Varmeziar, Armon
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Text mining was utilized to improve personalized tyrosine kinase inhibitor (TKI) therapy selection for chronic myeloid leukemia (CML) patients and identify long-term adverse events for which regular surveillance is recommended. TKI therapy remains the gold standard of current treatment and has greatly improved patient survival and outcome. However, the adverse events and side effects associated with TKIs negatively impact patient quality of life and therapy compliance. Because many patients must take TKIs indefinitely, understanding the long-term adverse event profile is critical for optimal clinical management. Given many TKI are relatively new, the long-term adverse event profiles are not yet fully understood. Text mining methods such as bag-of-words, k-means clustering, cross-domain literature-based discovery, and literature meta-analysis were conducted to predict and quantify the most relevant adverse events.
Three tiers of predicted adverse events were quantified through statistical analysis and machine learning. Tier 1 adverse events (top 1% of predictions) constituted hematological, glucose-related, iron-related, cardiovascular, and thyroid-related adverse events from TKI therapy. Tier 2 adverse events (top 5% of predictions) constituted kidney and inflammation related adverse events. Tier 3 adverse events (top 10% of predictions) constituted gastrointestinal, neuromuscular, or other adverse events like secondary malignancies. Based on study results, tier 1 conditions were recommended for regular surveillance; tier 2 for periodic surveillance with periodicity based on patient history; tier 3 for surveillance as a function of symptoms present. These insights provide a more proactive approach for optimal CML management and personalized TKI selection.
Sponsor
Date
2022-05-03
Extent
Resource Type
Text
Resource Subtype
Thesis