Thesis Defense | Shahul Alam | November 5, 2026 | 1pm
CBD and CPCB are proud to announce the following thesis defense:
Title: Interpretable and Scalable Machine Learning Frameworks for Spatial Transcriptomics Analysis
Shahul Alam
Thursday, November 5th @ 1PM ET
GHC 4405
For Zoom information, please reach out to Nicole Stenger
Committee:
Jian Ma (CMU)
Martin Zhang (CMU)
Maria Chikina (Pitt)
Xiao Wang (MIT)
Bonnie Berger (MIT)
Abstract: The rapid expansion of spatially resolved transcriptomics (SRT) technologies has created unprecedented opportunities to study the spatial organization of gene expression in health and disease. However, current computational approaches are constrained by tradeoffs between interpretability and scalability, limited incorporation of biological priors, and variability in data quality, all of which hinder the integrated analysis of SRT datasets. This Ph.D. dissertation aims to tackle these challenges through three aims, each centered on a novel machine learning framework. First, Popari, a probabilistic modeling framework, provides an interpretable approach to analyzing spatial variation across SRT samples. Second, Heimdall, a framework for evaluating tokenization strategies for single-cell transcriptomes, disentangles the effects of biological priors on downstream foundation model performance. Third, Aeromatik, an SRT foundation modeling framework, integrates single-cell RNA-seq and SRT data and introduces bespoke pretraining strategies to enable multiscale representation learning of tissues. Collectively, these approaches aim to bridge critical gaps in the spatial genomics field, enabling robust, interpretable, and scalable analyses of large-scale SRT datasets. This dissertation advances our understanding of spatial transcriptomic heterogeneity and its impact on cellular function in a wide range of biological contexts.