Bioinformatics Seminar
Time: 11AM
Venue: Hybrid
16 May 2023
Using machine learning to predict cancer outcomes...and to write code
Stephen PiccoloBrigham Young University
A key challenge of precision medicine is classifying patients who have a given disease into appropriate subtypes or treatment groups, rather than using a one-size-fits-all approach. For example, if we can identify characteristics that are common to patients who respond well to a particular treatment, clinicians might be able to use these patterns to assign new patients to the same treatment, potentially resulting in better responses, lower morbidity rates, and reduced costs. For 20+ years now, researchers have envisioned the potential to perform such classification tasks using transcriptomic data (in complement to clinical and demographic data). However, hundreds of classification algorithms are available, and each can be customized using hyperparameters, so it is difficult for researchers to optimize algorithm selection. Stephen will describe a large benchmark study that provides new insights on this topic. Additionally, he will describe findings from recent work on using ChatGPT as an aid for bioinformatics research tasks and as an educational resource for learning bioinformatics skills.