22 settembre 2026 ore 10:00
Laboratorio Antonio Amato - III piano (Palazzina Presidenza)
Via del Castro Laurenziano 9, Roma - Facoltà di Economia
Relatore: Alessandro Zito (Harvard University)
Titolo: Statistical methods for mutational processes in cancer
Abstract: Cancer cells acquire DNA mutations through many mutational processes, such as environmental exposures and dysregulated repair mechanisms. Each process consistently produces different mutation types at characteristic frequencies, referred to as its “mutational signature". The usual approach to infer these signatures consists of decomposing the matrix of mutation counts from a sample of patients using non-negative matrix factorization (NMF). However, working with aggregate counts ignores heterogeneous patterns of mutation rates across the genome and their relationship to tissue-specific characteristics. In this talk, I address these limitations by introducing the Poisson process factorization (PPF), a novel dimensionality reduction method that models mutation occurrences using inhomogeneous Poisson point processes. PPF generalizes the baseline NMF model by representing a patient's exposure to each signature as a locus-specific function of genomic covariates. This allows us to quantify the joint relationship between genomic features and mutational processes, revealing novel insights into breast cancer.
Bio: Alessandro Zito is a Postdoctoral research fellow in Biostatistics at the Harvard T.H. Chan School of Public Health and the Dana-Farber Cancer Institute, where he works under the supervision of Professor Jeffrey W. Miller and Professor Giovanni Parmigiani. He obtained his Ph.D. in Statistical Science at Duke University, with David B. Dunson as advisor. His research focuses on developing interpretable dimensionality reduction methods and stochastic processes to infer complex patterns in large-scale data, with applications to cancer genomics and ecology.
Tutti gli interessati sono invitati a partecipare.