Try tGPLVM out for visualization and processing scRNA-seq count data. The t-distributed noise means tGPLVM is robust to outliers, is able to go to higher dimensions than t-SNE, and does not suffer from mode collapse. Software available here: github.com/architverma1/t…
The Engelhardt Group develops statistical models and methods for high-dimensional biomedical data. In particular, we study human genetic variation, genomic regulation, single cell (spatial) transcriptomics, longitudinal cohort data, and electronic health care records (EHR) in order to identify mechanisms of human disorders and diseases. There are a wide range of projects in these areas in the group (see Research for more information), including:
- Spatiotemporal modeling of single cell transcriptomic data;
- Experimental design for single cell data and atlas design;
- Analysis of Perturb-seq data across time;
- Statistical models for longitudinal data in the Fragile Families Childhood Wellbeing Cohort;
- Time-series models of hospital patient state across time; and
- Reinforcement learning methods for safe policies for clinician-in-the-loop hospital settings.
Prof. Engelhardt is a PI in the Genotype-Tissue Expression (GTEx) consortium, within which our group is developing statistical models to analyze cis- and trans-eQTLs across cell types and the sexual dimorphism of transcription regulation.
Postdoc job opening: There is a postdoc opening in my group for a talented statistical geneticist, quantitative geneticist, or statistician who wants to study biomedical phenomena by developing data-driven statistical or machine learning methods. Please contact me if you might be that postdoc (bee-at-princeton-dot-edu).
Prospective graduate students: If you are interested in working in our group at Princeton for your PhD, you must be admitted to a graduate program at Princeton University such as Computer Science or Quantitative and Computational Biology.