About me

I am a Research Associate for Bayesian Statistics and Causal Inference at the School of Public Health, Imperial College London, working with Verena Zuber and Leonardo Bottolo. My research focuses on developing Bayesian latent variable models for high-dimensional molecular data, including gene expression and other omics data. I am particularly interested in latent factor models, biclustering, structured sparsity, clustering and scalable inference, with the aim of identifying meaningful latent structure in complex biological datasets. My current work focuses on Bayesian biclustering methods for heterogeneous molecular data, including multi-study and single-cell settings, as well as scalable variational inference for Bayesian latent representation models.

In 2023, I obtained my PhD at Vienna University of Economics and Business under the supervision of Sylvia Frühwirth-Schnatter and Alfred Stiassny. My dissertation, “Bayesian methods for unsupervised data analysis in application to data sets exhibiting non-Gaussianity,” focused on uncovering structure in complex data with minimal assumptions. It includes a review of infinite factorisation models, a latent factor model relaxing the Gaussian assumption, and flexible mixtures of factor analysers that enable automatic inference of both cluster structure and latent dimensionality. This work led to a recent publication in Bayesian Analysis.

Prior to academia, I worked for several years at the Research Department of Erste Group as a Quantitative Analyst, where I was responsible for macroeconometric forecasting and time series modelling.

My detailed CV can be found here: CV

Research Interests

• Bayesian latent variable and factor models
• Bayesian representation learning
• Biclustering and structured sparsity
• Mixture models and clustering methods
• Bayesian nonparametrics
• Statistical methods for high-dimensional molecular data
• Variational inference
• Time series models and forecasting

Current projects

Multi-study factor regression biclustering
A Bayesian latent factor model for high-dimensional molecular data collected across multiple studies. The model combines multi-study factor analysis, factor regression and biclustering within a unified framework, allowing shared biological structure to be separated from study-specific variation while identifying subsets of molecular features that are active only in subsets of samples. The accompanying inference framework is based on scalable variational approximations, together with post-processing procedures for assessing identification of the recovered shared latent factors. A working draft of the manuscript is available at: Multi-study factor regression biclustering

Bayesian biclustering for multi-subject single-cell transcriptomic count data
A Bayesian latent variable model for overdispersed single-cell RNA-seq count data with repeated measurements across individuals. The model introduces subject-specific activation of shared latent factors, allowing biological processes to be shared only across subsets of subjects while accounting for within-subject dependence and gene-specific overdispersion. Inference is carried out using variational Bayes with augmentation-based methods for tractable computation of all key model parameters.

Upcoming Talks

• Talk at MASAMB 2026 (Cambridge, September 2026)
• Invited talk at CMStatistics 2026 (Berlin, December 2026)

Publications

Papers and Preprints

• Dynamic Mixture of Finite Mixtures of Factor Analysers. M. Grushanina and S. Frühwirth-Schnatter (2025). Bayesian Analysis (advance publication).
  published paper | arXiv preprint

• A Review of Bayesian Methods for Infinite Factorisations.
  M. Grushanina (2023).
  arXiv preprint

Conference Proceedings

• Bayesian infinite factor models with non-Gaussian factors.
  M. Grushanina and S. Frühwirth-Schnatter (2021).
  In JSM Proceedings, International Society for Bayesian Analysis (ISBA) Section. Alexandria, VA: American Statistical Association. 396-415.
  conference paper

Discussions

• Discussion on Sparse Bayesian Factor Analysis When the Number of Factors Is Unknown by S. Frühwirth-Schnatter, D. Hosszejni, and H. F. Lopes.
  M. Grushanina (2025). Bayesian Analysis, 20(1), pp. 316-318.
  discussion

Presentations & Posters

• Poster at CSML Workshop 2026 (London, April 2026). Multi-study factor regression biclustering. Poster
• Talk at Research Computing Showcase Day 2026, Imperial College London (London, April 2026). Multi-study factor regression biclustering. Recording
• Seminar at EUSP (virtual, January 2025). Dynamic mixture of finite mixtures of factor analysers.
• Invited talk at CFE-CMStatistics 2024 (London, December 2024). Dynamic mixture of finite mixtures of factor analysers.
• Poster presentation at 2024 ISBA World Meeting (Venice, July 2024). Dynamic mixture of finite mixtures of factor analysers model with automatic inference on number of clusters and factors.
• Contributed talk at EWMES 2023 (Manchester, December 2023). Dynamic mixture of finite mixtures of factor analysers model with automatic inference on number of clusters and factors.
• Invited talk at IDWSDS 2023 (virtual, October 2023). Dynamic mixture of finite mixtures of factor analysers model with automatic inference on number of clusters and factors.
• Contributed talk at ESOBE 2023 (Glasgow, September 2023). Dynamic mixture of finite mixtures of factor analysers model with automatic inference on number of clusters and factors.
• Invited talk at CMStatistics 2022 (London, December 2022). Dynamic mixture of finite mixtures of factor analysers model with automatic inference on number of clusters and factors.
• Talk in young researcher session at ESOBE 2022 (Salzburg, September 2022). Dynamic mixture of finite mixtures of factor analysers model with automatic inference on number of clusters and factors.
• Poster presentation at 2022 ISBA World Meeting (Montreal, June 2022). Mixture of factor analysers and adaptive telescoping sampler for automatic inference on number of clusters and factors.
• Contributed talk at BNP 2022 Networking Event (Nicosia, April 2022). Mixture of factor analysers and adaptive telescoping sampler for automatic inference on number of clusters and factors.
• Seminar at CEBA (virtual, December 2021). Bayesian infinite factor models with non-Gaussian factors.
• Contributed talk at JSM 2021 (virtual, August 2021). Bayesian infinite factor models with non-Gaussian factors.
• Contributed talk at 2021 ISBA World Meeting (virtual, July 2021). Bayesian infinite factor models with non-Gaussian factors.

Selective Work Experience

• 2024-present: Research Associate for Bayesian Statistics and Causal Inference, School of Public Health, Imperial College London (London, UK)
• 2014-2024: Economist/quantitative analyst at Research Department of Erste Group (Vienna, Austria).
• 2007-2011: Data analyst at Strategic Risk Management Department of Erste Group (Vienna, Austria).

Miscellanea

I love music and have studied classical piano and vocals for some years. I have also participated in several summer courses in jazz and creative music/songwriting at Guildhall School of Music and Drama as well as had some amateur performances at various events.

Contact me

m.grushanina at imperial.ac.uk
margarita.grushanina at gmail.com