Agenda

Speaker: Francesco Denti, Università di Padova

Link Zoom: https://unive.zoom.us/j/85153268624?pwd=MzBhdlA2M1B2dThJQ2Y5T0EwUE5PZz09
Meeting ID: 851 5326 8624
Passcode: SanMarco2

Abstract:
In this talk, we discuss two Bayesian nonparametric models designed to address the challenges posed by large, complex biological datasets generated by state-of-the-art imaging technologies. The first application focuses on a MALDI-mass spectrometry imaging dataset, which quantifies the abundance of numerous specific molecules across multiple locations within a biological tissue sample. We propose a model for nested, separate exchangeable data inducing a biclustering to simultaneously group spatial locations with similar abundance profiles and molecules with similar expression patterns. A hidden Markov random field prior is incorporated to enable precise image segmentation, ensuring that clusters correspond to biologically meaningful regions. The resulting biclustering structure facilitates the detection of molecular activation patterns and provides interpretable segmentations of the analyzed image. The second model addresses calcium imaging data, which record the activity of individual neurons over time in freely moving animals. We develop a spatiotemporal mixture model to identify co-activating neurons, detecting groups of cells with a similar firing activity over time. This is achieved through a multivariate time-series framework that detects spikes in calcium traces and captures recurring temporal patterns. We employ a Dependent Dirichlet Process to incorporate information on the anatomical proximity between neurons. Simultaneously, spikes' amplitudes are segmented using a Dirichlet process, providing clusters based on their magnitudes.

Bio Sketch:
Francesco is an Assistant Professor (RtdB) at the Department of Statistics of University of Padua.
Previously, he was Assistant Professor (RtdA) at the Department of Statistics of Università Cattolica del Sacro Cuore - Milan. Prior to that, he spent two years as a Postdoctoral
Scholar at the Department of Statistics of Univerisity of California - Irvine, working under the supervision of Professor Babak Shahbaba.
He holds a Ph.D. in Statistics at University of Milan - Bicocca and Università della Svizzera italiana (joint program, awarded with honors) - supervised by with Professor Antonietta Mira and Professor Michele Guindani.
His research interests mainly relate to Bayesian mixtures, Bayesian nonparametrics, model-based clustering, shrinkage priors, and dimensionality reduction.