ALLSTAR: inference of reliAble causaL ruLes between Somatic muTAtions and canceR phenotypes

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Tumor genomes exhibit extreme heterogeneity, complicating the identification of driver mutations.
• Existing tools correlate somatic mutations with cancer phenotypes but do not guarantee causality.

Purpose of the Study:

• To develop a novel computational tool, ALLSTAR, for inferring reliable causal relations between somatic mutations and cancer phenotypes.
• To identify combinations of somatic mutations with the highest impact on cancer phenotypes.

Main Methods:

• Developed a branch-and-bound algorithm to address the NP-hard computational problem.
• Integrated protein-protein interaction networks and novel bounds for efficient search space pruning.
• Implemented rigorous multiple hypothesis testing correction.

Main Results:

• ALLSTAR successfully infers reliable causal relations from synthetic and real cancer data.
• The tool identifies known cancer-associated somatic mutations and novel biologically relevant relationships.
• Demonstrated the tool's ability to handle large cancer cohorts and complex genomic data.

Conclusions:

• ALLSTAR provides a robust method for uncovering causal links between somatic mutations and cancer phenotypes.
• The identified causal rules offer insights into cancer biology and potential therapeutic targets.
• This approach advances the understanding of tumor heterogeneity and its clinical implications.