Applying unmixing to gene expression data for tumor phylogeny inference

Area of Science:

• Computational biology
• Cancer research
• Genomics

Background:

• Most human cancers fall into a limited number of sub-types, each with a characteristic mutation sequence.
• Identifying common cancer sub-types is crucial for developing diagnostics and therapeutics.
• Current phylogenetic methods for tumor progression face limitations in assaying intra-tumor heterogeneity or averaged tumor states.

Purpose of the Study:

• To develop and validate an unmixing method for cancer phylogenetics.
• To infer recurring cell states and evolutionary relationships within tumors.
• To leverage advantages of both single-cell and tissue-wide approaches.

Main Methods:

• Applied "unmixing" methods to microarray data of tumor populations.
• Inferred recurring cell states from complex datasets.
• Used inferred mixtures of states to identify evolutionary relationships among tumor cells.

Main Results:

• The unmixing method accurately separates cell states and infers phylogenetic relationships on simulated data.
• Applied to lung cancer data, the method identified common tumor cell states.
• Suggested evolutionary relationships among lung tumor cells align with current understanding.

Conclusions:

• Unmixing methods integrate intra-tumor heterogeneity and large probe sets for robust tumor phylogeny inference.
• This approach enables detailed and accurate portraits of common cancer sub-types and their developmental mechanisms.
• Inferred tumor phylogenies hold potential for discovering/diagnosing cancer sub-types and identifying therapeutic targets.