Biostat Student Seminar Series

Speaker: Cathy Xue
Title: Robust discovery of mutational signatures using power posteriors
Abstract:
Mutational signatures are distinctive patterns of mutations resulting from carcinogenic molecular processes, such as UV radiation, molecular effects of chemical agents, and defective DNA repair mechanisms. Non-negative matrix factorization (NMF) models have been used to discover mutational signatures and deconvolve their respective contributions in individual tumors from sequencing data. However, any assumed model is only a rough approximation to reality, and as a consequence, the results are sometimes misleading and irreproducible. We propose an alternative approach to mutational signature inference that, by leveraging the power posterior, is robust to using an approximate model and, by using a novel sparsity-inducing prior, automatically infers the number of signatures. We demonstrate the robustness and accuracy of our approach on simulated data and real data from the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium.

Speaker: Yi-Ting Tsai
Title: Predicting facility-based delivery in Zanzibar: the vulnerability of machine learning algorithms to adversarial attacks
Abstract:
Background: Community health worker (CHW)-led maternal health programs have contributed to increased facility-based deliveries and decreased maternal mortality in sub-Saharan Africa. The recent adoption of mobile devices in these programs provides an opportunity for real-time implementation of machine learning predictive models to identify women most at risk for home-based delivery. However, it is possible that falsified data could be entered into the model to get a specific prediction result – known as an “adversarial attack”. The goal of this project is to evaluate the algorithm’s vulnerability to adversarial attacks.
Methods: The dataset used in this research is from the Uzazi Salama (“Safer Deliveries”) program, which operated between 2016 and 2019 in Zanzibar. We used LASSO regularized logistic regression to develop the prediction model. We used “One-At-a- Time (OAT)” adversarial attacks across four different types of input variables: binary – access to electricity at home, categorical – previous delivery location, ordinal – educational level, and continuous – gestational age. We evaluated the percent of predicted classifications that change due to these adversarial attacks.
Results: Manipulating input variables affected prediction results. The variable with the greatest vulnerability was previous delivery location, with 55.65% of predicted classifications changing when applying adversarial attacks from previously delivered at a facility to previously delivered at home, and 37.63% of predicted classifications changing when applying adversarial attacks from previously delivered at home to previously delivered at a facility.
Conclusion: This project investigates the vulnerability of an algorithm to predict facility-based delivery when facing adversarial attacks. By understanding the effect of adversarial attacks, programs can implement data monitoring strategies to assess for and deter these manipulations. Ensuring fidelity in algorithm deployment secures that CHWs target those women who are actually at high risk of delivering at home.

——————————————————–
Upcoming seminars:
We have booked FXB G11 for the following dates for the remaining seminars this semester:

Thursday, December 8

We are still looking for speakers for these dates. Please reach out if interested! As a reminder, most talks run around 35–45 minutes, with the remaining seminar time left over for student questions and discussion. Two students can also split a seminar slot, in which case the individual talks are typically briefer (around 20 minutes each).

——————————————————–
What is the Biostatistics PhD/Master’s Student Seminar Series?
The series is a casual and relaxing place for Biostatistics students to informally gather roughly every other week to share research, connect with other students, give practice talks, and see what everyone else is working on over lunch, provided by the department. Some seminars are directly related to research students are doing for their degrees, but others are more generally about graduate school and the practice of statistics.

Please note that the seminars are open only to PhD students, Master’s students and postdocs in biostatistics.

————————————————————-

Best,
Your 2022-2023 Biostatistics Student Seminar Organizing Committee
Izzy Grabski and Cathy Xue

(student-only event)