VICTree
Variational inference for joint reconstruction of clonal trees and copy-number evolution from single-cell DNA sequencing data.
Python · variational inference · HMM
Research software
Italy · Portfolio
Vittorio Zampinetti PhD
Machine Learning · Software Engineering · Maker / Geek
Researcher and engineer interested in probabilistic models, algorithms and the evolutionary history of cancer cells. I completed my PhD between Politecnico di Torino and KTH, working with Jens Lagergren. I enjoy turning mathematical ideas into reliable, open software.
Selected work & open source
projects
Research implementations, teaching material and software experiments.
TreeSampling
Efficient sampling of random spanning arborescences, including directed graphs with low conductance where conventional walks struggle.
Python · random walks · dynamic programming
Open source
Cellmates
Single-cell phylogeny reconstruction with evolutionary distances derived directly from copy-number profiles.
Python · phylogenetics · copy numbers
Research software
Sparse Rooted Neighbor Joining
A scalable rooted phylogenetic reconstruction method designed for single-cell copy-number data and its known diploid root.
Python · phylogenetics · neighbor joining
Research software
Statistical Models in R
Open teaching material introducing modern R for statistical analysis, written for the Statistical Models MSc course at Politecnico di Torino.
R · statistics · Quarto
Open textbook
Peer-reviewed work & preprints
publications
Probabilistic modelling and algorithms for computational biology.
Cellmates—Single-Cell Phylogeny Reconstruction with Proper Evolutionary Distances from Copy Numbers
bioRxiv preprint
Cellmates addresses single-cell tumour phylogeny reconstruction by deriving proper evolutionary distances from copy-number profiles and using those distances in phylogenetic reconstruction. The work was also presented at the 2026 Bertinoro Computational Biology Meeting.
Code for Cellmates—Single-Cell Phylogeny Reconstruction with Proper Evolutionary Distances from Copy NumbersBibTeX
@article{zampinetti2026cellmates, author = {Zampinetti, Vittorio and Melin, Harald and Abdolhamdi, Marzie and Lagergren, Jens}, title = {Cellmates: a method for single-cell phylogeny reconstruction with proper evolutionary distances from copy numbers}, journal = {bioRxiv}, year = {2026} }Scalable and Robust Phylogenetic Tree Reconstruction from Copy-Number Data with Sparse Rooted Neighbor Joining
bioRxiv · Cold Spring Harbor Laboratory
Rooted Neighbor Joining variants offer a principled way to use the known diploid root when reconstructing phylogenies from copy-number data. Sparse Rooted Neighbor Joining extends this advantage to datasets whose size places the full distance matrix out of reach, reducing running time with only a minor loss of accuracy in the evaluated settings.
DOI for Scalable and Robust Phylogenetic Tree Reconstruction from Copy-Number Data with Sparse Rooted Neighbor Joining PDF for Scalable and Robust Phylogenetic Tree Reconstruction from Copy-Number Data with Sparse Rooted Neighbor JoiningBibTeX
@article{zampinetti2026srnj, author = {Zampinetti, Vittorio and Melin, Harald and Hallin, Alva and Lagergren, Jens}, title = {Scalable and robust phylogenetic tree reconstruction from copy-number data with Sparse Rooted Neighbor Joining}, journal = {bioRxiv}, year = {2026}, doi = {10.64898/2026.07.30.739152} }Sampling Random Spanning Arborescences in Graphs with Low Conductance
Statistics & Probability Letters · Elsevier
Sampling random spanning arborescences in directed graphs is important for network analysis, optimisation and machine learning. Inspired by Wilson’s algorithm, this work uses dynamic programming to compute random-walk probabilities, avoiding inefficient walk simulations and unstable Laplacian determinant calculations. The approach improves efficiency and sampling quality, particularly on low-conductance graphs.
DOI for Sampling Random Spanning Arborescences in Graphs with Low Conductance Code for Sampling Random Spanning Arborescences in Graphs with Low ConductanceBibTeX
@article{treesampling2025, author = {Zampinetti, Vittorio and Melin, Harald and Lagergren, Jens}, title = {Sampling random spanning arborescences in graphs with low conductance}, journal = {Statistics & Probability Letters}, volume = {226}, pages = {110481}, year = {2025}, doi = {10.1016/j.spl.2025.110481} }VICTree: A Variational Inference Method for Clonal Tree Reconstruction
Research in Computational Molecular Biology (RECOMB) · Springer Nature Switzerland
Clonal tree inference brings crucial insights to the analysis of tumour heterogeneity and cancer evolution. VICTree is a variational-inference algorithm for joint Bayesian inference of clonal trees, built on a novel tree-structured mixture hidden Markov model. Experiments on simulated data and samples of multiple myeloma and breast cancer evaluate clustering, tree reconstruction, copy-number evolution and model selection.
PDF for VICTree: A Variational Inference Method for Clonal Tree Reconstruction Code for VICTree: A Variational Inference Method for Clonal Tree ReconstructionBibTeX
@inproceedings{victree2024, author = {Melin, Harald and Zampinetti, Vittorio and McPherson, Andrew and Lagergren, Jens}, title = {VICTree - A Variational Inference Method for Clonal Tree Reconstruction}, booktitle = {Research in Computational Molecular Biology}, pages = {429--433}, year = {2024}, publisher = {Springer Nature Switzerland} }
Notes from research and beyond
blog
Short updates on publications, conferences and technical work.
- Read
Single-Cell Tumor Phylogenetics
A short note on defending my PhD thesis and closing an important research chapter.
- Read
Cellmates at Bertinoro 2026
Presenting the final Cellmates results at the Bertinoro Computational Biology Meeting.
- Read
Sampling Arborescences, Published
Our article on sampling random spanning arborescences is now available as open access.
- Read
Work in Progress at Bertinoro
Early Cellmates work and a week of computational biology talks in Bertinoro.
- Read
Presenting VICTree in Bari
A presentation on variational inference for clonal tree reconstruction at SIS 2024.
- Read
VICTree at RECOMB 2024
Presenting our variational-inference work at RECOMB in Boston.