Methods and Analyses – Sottoriva Lab

MIDAA

Multiomics Integration via Deep Archetypal Analysis

MIDAA is a package designed for performing Deep Archetypal Analysis on multiomics data.

Reference: Milite 2025, Genome Biology ⌥ Code on GitHub

NEUROVELO

NeuroVelo: interpretable learning of cellular dynamics

NeuroVelo is a physics-based method for interpretable learning of cellular dynamics. Implemented in Python3 and PyTorch, the model estimates the velocity field and the genes that drive the splicing dynamics.

Reference: Kouadri Boudjelthia 2026, Cell Reports Methods ⌥ Code on GitHub

EPICC ANALYSIS

EPICC papers data analysis pipelines

Data analysis pipelines for the EPICC studies on the co-evolution of the genome and epigenome and on phenotypic plasticity in colorectal cancer.

Reference: Heide, Househam et al. 2022, Nature; Househam, Heide et al. 2022, Nature ⌥ Epigenome code ⌥ RNA code

EPICC SIMULATIONS

EPICC simulations and inference

Simulation and inference framework for the EPICC second paper on phenotypic plasticity and genetic control in colorectal cancer evolution.

Reference: Househam, Heide et al. 2022, Nature ⌥ Inference code ⌥ MLLPT code

MOBSTER

Subclonal reconstruction combining evolutionary theory with machine learning

A novel approach for model-based tumour subclonal reconstruction that combines machine learning with theoretical population genetics, validated on public whole-genome sequencing data from 2,606 samples.

Reference: Caravagna et al. 2020, Nature Genetics ⌥ Code on GitHub

MCMC – MUTATIONAL DISTANCES

Inference of single-cell microscopic parameters from multi-sampling genomic data

Inference of de-coupled single-cell microscopic parameters — including mutation rates per division and cell death rates — from multi-sampling genomic data using Markov chain Monte Carlo methods.

Reference: Werner et al. 2020, Nature Communications ⌥ Code on GitHub

CHESS

Spatial model of tumour growth and genomic data generation

A spatial model of tumour growth that simulates different sampling strategies and the generation of genomic data, enabling the study of how spatial constraints shape clonal evolution.

Reference: Chkhaidze et al. 2019, PLOS Computational Biology ⌥ Code on GitHub

REVOLVER

Detecting repeated evolutionary trajectories in cancer using Transfer Learning

REVOLVER (Repeated EVOLution in cancER) uses Transfer Learning to analyse and compare phylogenetic trees from multi-region sequencing studies, identifying hidden patterns of repeated evolutionary trajectories that correlate with prognosis.

Reference: Caravagna et al. 2018, Nature Methods ⌥ Code on GitHub

QUANTIFYING SELECTION

Measuring neutral evolution and subclonal selection from patient genomic data

A set of tools to measure neutral evolution and quantify subclonal selection, using either a frequentist approach or an Approximate Bayesian Computation approach for model selection.

Reference: Williams et al. 2016, Nature Genetics; Williams et al. 2018, Nature Genetics ⌥ Code