The Grabski Lab develops statistical and machine learning methods to uncover biological mechanisms from large-scale genomic and imaging data. Advances in single-cell and spatial genomics, high-throughput perturbation screens, and imaging technologies are generating increasingly rich views of biological systems, yet translating these complex measurements into mechanistic insights remains a fundamental computational challenge. Our research aims to address this gap by developing interpretable computational frameworks that harness these expanding data resources to reveal the molecular programs shaping biological states and phenotypes in health and disease.
1.
Perturbation-based interpretation of cellular states
Systematic perturbation experiments provide a powerful way to uncover the causal mechanisms underlying biological variation. Our previous work introduced RNA fingerprinting, a computational framework that learns transcriptional signatures from large-scale perturbation screens. By matching these signatures to new datasets, we can interpret observed cellular states and responses in terms of their underlying molecular mechanisms. We aim to extend this approach to more complex biological systems, developing statistical models that capture how perturbations reshape cellular states and collective behaviors across tissues and organisms. These models will provide a foundation for interpreting complex biological phenotypes through the lens of experimentally defined perturbation responses. We will further extend these frameworks to capture how perturbation responses unfold over time, such as during development, and across space, leveraging emerging spatial perturbation technologies to reveal how molecular programs and cellular interactions shape tissue organization.
2.
Understanding the consequences of genetic and somatic variation
Genetic variation is a key driver of biological diversity and disease, yet understanding how individual variants influence cellular states and tissue function remains a major challenge. Unlike experimental perturbations, naturally occurring genetic alterations often have subtle, context-dependent effects, and somatic mutations can give rise to complex patterns of cellular heterogeneity and clonal evolution. Our lab develops statistical and machine learning methods to link inherited and somatic genetic variation to its molecular and cellular consequences, leveraging large-scale single-cell genomic datasets to uncover the programs associated with specific genetic alterations. We are particularly interested in understanding how somatic mutations shape both the states of mutant cells and their surrounding microenvironment, and in developing models that capture how successive mutations influence biological phenotypes along evolving clonal lineages. Together, these approaches will provide new ways to understand how genetic variation contributes to disease, aging, and tissue dysfunction.
3.
Connecting molecular mechanisms to cellular and tissue morphology
Imaging provides a unique window into biological function, capturing phenotypes ranging from cellular morphology to tissue architecture and histopathology. These measurements are also central to many clinical settings, where morphological features can provide important insights into disease state and progression. Emerging technologies, including high-throughput microscopy, optical pooled screening, and spatial profiling, are generating increasingly rich imaging datasets and opening new opportunities to uncover the molecular mechanisms underlying these complex phenotypes. We aim to develop statistical and machine learning methods to learn interpretable representations of image-based phenotypes and establish relationships between morphological patterns and their underlying molecular programs. By leveraging systematic perturbation experiments alongside increasingly rich spatial and molecular measurements, we seek to understand how biological processes give rise to distinct morphological phenotypes and use these relationships to interpret images in new biological and clinical contexts, including histopathology.
