Ozgun Babur
University of Massachusetts-Boston, MA·Computer Science·Early-career
Modern laboratories can measure thousands of molecules in a cell at once, from proteins to their chemical modifications to the transcripts that encode them. Turning those huge tables of numbers into an account of what caused what is still largely manual, slow and a bottleneck on discovery.
- Fields
Computer Science
Biochemistry
Genetics
- Topics
causal modeling
CausalPath
multi-omics
- Grants
- 1 known grant, through Aug 2030
What they work on
- Extend the CausalPath framework, which interprets multi-omic data through causal biological networks, to single-cell omics trajectories and to model organisms via homology-based gene mapping.
- Improve explanations of differential phosphorylation by integrating kinase substrate sequence preference motifs, the patterns kinases recognize in the proteins they modify, into the causal models.
- Infer the functional consequences of individual protein phosphorylation events systematically by jointly analyzing large-scale phosphoproteomic and transcriptomic datasets.
- Develop causality-aware graph neural networks with new convolution and attention mechanisms that use causal graph structure to predict phenotypes from molecular profiles more accurately.
More about the research
This project builds computational infrastructure to automate the construction of causal models, networks that propose which molecular events drive which others, so that cause-and-effect reasoning can be applied at the scale of today's data.
Why it matters
Biologists studying cancer, development and drug response routinely collect multi-omic data they cannot fully interpret. Automated causal modeling could make that interpretation routine rather than bespoke, speeding the identification of which molecular events drive a cellular change and extending the approach to single-cell data and model organisms. The resulting open methods and new courses could broaden who can analyze such datasets.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Mark Chance
Case Western Reserve University, OH·Biochemistry
Provide proteomic and small-molecule analysis to cancer investigators, including protein identification, post-translational modification characterization, quantitative proteomics, protein interactome mapping and metabolomics, across collaborative and high-throughput service modes.
Ritambhara Singh
Brown University, RI·Computer Science·Early-career
Predict gene expression by integrating histone modification profiles, 3D genome conformation, DNA sequence and single-nucleotide polymorphisms in deep learning models, including graph neural networks that treat genome folding as a network of contacts between distant DNA regions.
Yufei Huang
University of Pittsburgh, PA·Computer Science
Develop a contrastive learning-based geometric graph attention model that learns gene expression patterns aware of tissue shape from single-cell spatial transcriptomics data, so that the patterns match the tissue structure seen in histology.
The school
University of Massachusetts-Boston: 3 labs on PI Pages, $75M a year on research, 108 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.