Related Experiment Video
Updated: Jun 24, 2025
08:16
Comparative Lesions Analysis Through a Targeted Sequencing Approach
Published on: November 5, 2019
6.7K
Crowd-sourced benchmarking of single-sample tumor subclonal reconstruction
Adriana Salcedo1,2,3,4,5, Maxime Tarabichi6,7,8, Alex Buchanan9
1Department of Human Genetics, University of California, Los Angeles, CA, USA. ASalcedo@mednet.ucla.edu.
Nature Biotechnology
|June 11, 2024
Summary
Subclonal reconstruction algorithms analyze tumor evolution from DNA sequencing data. No single algorithm excelled across all tasks, indicating a need for improved methods to accurately assess cancer progression.
Area of Science:
- Genomics
- Computational Biology
- Cancer Research
Background:
- Subclonal reconstruction algorithms are crucial for understanding tumor evolution using bulk DNA sequencing data.
- These algorithms help assess cancer initiation, progression, and response to treatment.
Purpose of the Study:
- To benchmark existing subclonal reconstruction algorithms using a large-scale challenge.
- To identify factors influencing algorithm performance and highlight areas for improvement.
Main Methods:
- The International Cancer Genome Consortium-The Cancer Genome Atlas (ICGC-TCGA) DREAM Somatic Mutation Calling Tumor Heterogeneity and Evolution Challenge was established.
- 31 subclonal reconstruction algorithms were benchmarked on 51 simulated tumors using cloud computing.
- Algorithms were evaluated on seven independent tasks, resulting in over 12,000 runs.
Main Results:
- Algorithm choice significantly impacted performance more than tumor characteristics.
- Purity-adjusted read depth, copy-number state, and read mappability correlated with algorithm performance.
- No single algorithm outperformed others across all seven tasks; ensemble methods did not surpass the best individual algorithms.
Conclusions:
- There is a critical need for developing improved subclonal reconstruction algorithms.
- Further research should focus on understanding determinants of accuracy to enhance tumor evolution analysis.
- Containerized methods, code, and datasets are publicly available to facilitate future research.
