CAPRI: efficient inference of cancer progression models from cross-sectional data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Cancer genomic datasets, such as TCGA, provide cross-sectional measurements at diagnosis.
• Inferring cancer progression models from this data is crucial for understanding disease evolution.
• These models, represented as directed acyclic graphs (DAGs), map gene mutation selectivities.

Purpose of the Study:

• To develop a novel inference algorithm for reconstructing cancer progression models.
• To improve the accuracy and efficiency of inferring gene mutation selectivities from genomic data.
• To provide a tool that aids in patient stratification and personalized therapy.

Main Methods:

• Developed the CAncer PRogression Inference (CAPRI) algorithm.
• Utilized a scoring method based on Suppes' probabilistic theory.
• Incorporated bootstrap and maximum likelihood inference techniques.

Main Results:

• CAPRI demonstrates high accuracy and efficiency, outperforming state-of-the-art algorithms.
• The algorithm exhibits robust performance with noisy data and limited sample sizes.
• CAPRI successfully reconstructs confluent trajectories and uncovers selectivity patterns, as shown in a Chronic Myeloid Leukemia study.

Conclusions:

• CAPRI is an effective tool for cancer progression model reconstruction.
• The algorithm's performance and robustness offer significant advantages over existing methods.
• CAPRI has the potential to advance cancer research and clinical applications.