# Computational strategies in tumor phylogenetics: evaluating multimodal integration and methodological trade-offs across study designs

## TL;DR

This paper reviews and benchmarks computational tools for understanding tumor evolution, highlighting the need for multimodal data integration and new frameworks to improve cancer treatment strategies.

## Contribution

The paper introduces a novel spatiotemporal framework and benchmarks over 20 tools to guide future methodological improvements in tumor phylogenetics.

## Key findings

- Multiomics integration improves phylogenetic inference but struggles with mutation ordering and polyclonal detection.
- A spatiotemporal framework links phylogenetic branch lengths with spatial transcriptomic gradients.
- Benchmarking reveals key limitations in current tools, emphasizing the need for multimodal and scalable approaches.

## Abstract

Tumor clonal evolution represents a dynamic ecosystem underpinned by genetic alterations and Darwinian selection, posing major challenges due to intratumoral heterogeneity and therapy resistance. Although computational methods have advanced significantly, current tools often focus on single data modalities, leaving important gaps in modeling spatial and non-genetic evolution. This review systematically surveys and assesses algorithmic progress across diverse study designs to identify key limitations and future directions.

We systematically evaluate over 20 computational tools across four study designs—cross-sectional, regional bulk, single-cell, and lineage tracing—and perform benchmarking of seven comparable tools. Multiomics integration approaches are shown to improve phylogenetic inference, yet challenges remain in mutation ordering and polyclonal detection. A novel spatiotemporal framework is proposed to link phylogenetic branch lengths with spatial transcriptomic gradients. Future efforts should prioritize multimodal data integration, scalable computational architectures, and clinically applicable models to translate evolutionary insights into precision oncology.

This review provides a comprehensive survey and benchmarking of existing methods. The code and data used to generate the benchmarking figures and results are available at https://github.com/zlsys3/review_benchmark/tree/main/figure.